Home

Benchtop System Users Manual

1. S Orange Yellow Pw RGBR Mean Spectrum Crop inte New Cube Mean Sangard Deviation Heen First Genivstive Mean Second Der atis Send To Clipboard MENU REVEALED WITH RIGHT CLICK Spectrogram A Y 4500 S600 2700 1800 900 j il 400 FUU 600 FOU 800 SOO Contrats a duo Update LConkrast Presets True Color Color Infrared a 040 0 Red J 404 6005 63 5805 FH ib at Green 41S 6985 69 SES So 4625 AE 460 0 Elite 404 0085 73 5005 994 4025 Fig 6 1 A small ROI on one of the red candies has been selected and a right click has revealed the popup selection menu 6 4 Region Of Interest ROI Tools 49 Benchtop System Users Manual Release 4 3 accessing the Workbench tab from File Preferences menu Q Preferences Selection Transparency Floodfill petauit mage Open Cubes to Warnings Warn on Unsaved Last Scan The use of the floodfill tool depends on an adjustable tolerance parameter Floodfill operates on a representation of the datacube scaled from zero to one in each band It calculates the Euclidean distance or SAM angle in spectral space between the clicked pixel and all contiguous pixels and expands the selection until the selected area contains all of the contiguous pixels for which the spectral distance to the clicked pixel is less than the se
2. INTERFACE CONNECTION ON PIKA Using the provided interface cable connect your Pika imaging spectrometer to the appropriate port on your computer INTERFACE CONNECTION ON DESKTOP COMPUTER When using a Pika II camera with a GigE interface the camera should be plugged directly into the computer s network adapter not connected to the computer via a local area network hub or switch 3 2 2 Installing Calibration Values During initial setup you will input the calibration numbers for your Pika imaging spectrometer into SpectrononPro software The calibration numbers provide the information needed for SpectrononPro to provide the correct wave length values for the measured spectral bands Your Pika s calibration numbers are provided on a sheet that comes with your imaging spectrometer You only need to do this during initial setup or after an instrument re calibration If you cannot locate your calibration numbers contact Resonon at ProductSupport resonon com Connect your Pika imaging spectrometer to your computer Launch SpectrononPro by double clicking on the Spec trononPro icon E or navigate to it from your Start menu 3 2 Setup of Hyperspectral Imager 17 Benchtop System Users Manual Release 4 3 From the Main menu select File Preferences This will reveal the Preferences menu Select the tab for your Pika imaging spectrometer Data cube Ima spectri Open Datacube Open Spec
3. INTERFACES SPIN is also valid This line is not needed if you just want default behaviour self my float interfaceType INTERFACES SLIDER def action self my product self my float value 2 5 more plugin logic here 10 4 3 Specint class SpecInt SpecNumber Creates a slider or spinner for setting an int def init self label minval maxval stepsize 1 defaultValue None Using a SpecInt is just like using a SpecFloat See SpecFloat for an example 10 4 Specs 103 Benchtop System Users Manual Release 4 3 10 4 4 SpecChoice class SpecChoice Spec Creates a combo box depending allowing selection of a value from a list of choices def init self label values None defaultValue None A Spec choice allows the user to choose an option from a combo box It expects a list of strings as possible choices It s value member will be one of the strings in the list Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecChoice class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my_options SpecChoice label option values choice one choice two choice three defaultValue choice self my_options help my mouse hover help action self 1f self my_options value choice one pass action for choice one elif s
4. def getBand self bandnumber asFloat False Return band at given band number params bandnumber band number between 0 and bands 1 integer asFloat scale the datacube from 0 1 bool recurns band the image at the band number 2d array always lines samples def getFrame self line asFloat False asBIP False def 110 return frame at given line number params line the line number first 0 of a frame in the cube int asFloat scale the datacube from 0 1 bool returns frame the spectrum of each pixel in a line of the image 2D array always samples bands appendFrame self frame append frame to cube params frame the content of a frame of the same samples bands dimensions as the cube 2d array Chapter 10 Custom Data Analysis Writing Plugins def def def def def def Benchtop System Users Manual Release 4 3 appends a frame and increases number of lines in the cube by one appendBandWithName self band bandname None append band to cube params band the content of a band of the same lines samples dimensions as the cube 2d array appends a band and increases number of bands in the cube by one appendBandWithWavelength self band wave None append band to cube params band the content of a band of the same lines samples dimensions as the cube 2d array keywords writeheader default True if
5. Benchtop System Users Manual Release 4 3 SBsoerMmeel 0 inf x y Drange Yellow PW s RGE Resource Tree Spectrogram X i m Cubes M g DJ orange relow Fwos bip GRAB THE INSPECTOR TOOL HERE 4500 Orange Yellow Pia Rig Spectra w00 Calibration c 2700 900 ine doo 500 600 800 OO 1000 Corrales info Ja Cube CrDocumenks and Auto Update andi Dessbopi AG Msi Oren mm ge Tells PT bip Presets Render Too AGE Class retum an RGB S layer image from True coke 3 given bands of cube n Filter Tool Contrast Class Stretch Color Infrared Stretch the brightresss values between bio give percentage 3 540 0 amounts Red d Dat 414 6905 69 5805 94 4625 di 475 1 550 0 ine 255 Green J Ra 404 6955 695 5B05 994 4625 PIXEL LOCATION uu Blue PIXEL S RED GREEN AND BLUE BRIGHTNESS 44 6885 699 5005 094 4625 Click on other pixels to see the spectra from other pixels click and hold while dragging the inspector tool to update the plot panel continuously The red green and blue vertical lines in the spectral plot indicate the hyperspectral wavelength bands used to generate the current image 6 4 Region Of Interest ROI Tools It is often useful to consider a group of pixels within the image The ROI tools enable this capability and provide
6. Brightness Cube MCARI Render Tool Scalar to color map Class SingleBandColormap Return single band rendering using chosen colormap Filter Tool Contrast Class Stretch Stretch the brightness values between two give percentage amounts Applied render plugin Scalar to color map to datacube MCART After generating a colormap the user can change the color scheme from the Colormap menu on the Controls tab It may be helpful to toggle the Contrast Enhancement feature Experiment with different colormaps normalization and contrast settings all available on the Controls tab Note If an HVI is selected that requires spectral bands that do not exist in your datacube the warning shown below will appear As described in the warning clicking will generate an image using the closest bands available but the resulting image will often be a poor approximation to a true index mapping 96 Chapter 9 Advanced Data Analysis 3 Hyperspectral Vegetation Indices Benchtop System Users Manual Release 4 3 Wavelength Warning The datacube does not contain the necessary wavelength band The closest band has been used but the results may be inaccurate 9 2 Generating HVI maps in Spectronon 97 Benchtop System Users Manual Release 4 3 98 Chapter 9 Advanced Data Analysis 3 Hyperspectral Vegetation Indices CUSTOM DATA ANALYSIS WRITING PLUGINS Spectronon has the ca
7. Datacube New Cube Correct Correct From Measured Reference The Correct From Measured Reference option allows you to scale your datacube to a known reference source This option is useful for obtaining reflectance measurements during field deployments Since this tool does NOT account for spatial variability such as vignetting and slit variations the data should be converted to radiance before using this plugin The radiance conversion plugin is described below To utilize the Correct From Measured Reference tool it is best if there is a reflectance reference target located within your hyperspectral image First use one of the ROI i tools marquee or lasso or and generate a mean spectrum Then select Datacube New Cube Correct Correct From Measured Reference Use the Measured Reflectivity button to select the tab delimited file of measured reflectivty This will generate a new datacube scaled to brightness values of your Correction Spectrum If you click Float Mode the brightness values will be scaled to 1 Otherwise the data will be scaled to the maximum bit level of your datacube Datacube New Cube Correct Correct From Spectrally Flat Reference The Correct From Spectrally Flat Reference option allows you to scale your datacube to a spectrally flat reference source This option is useful for obtaining reflectance measurements during field deployments Since this tool
8. blue USB 3 0 port Plug the mini USB connector end into the stage motor 14 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 MiNI U SB CONNECTION NEXT COMNECT THE POWER TO YOUR STAGE USING THE PROVIDED DG POWER SUPPLY CONNECT F PLUG INTO 120 V MOTOR ON STAGE Screw the posts onto the bottom of your Pika imaging spectrometer The studs in the posts are 4 20 Install both posts in neighboring mounting holes on the bottom of the imaging spectrometer 3 1 Assembly of Stage Tower and Lighting 15 Benchtop System Users Manual Release 4 3 Insert the posts into the adjustable post holder on the tower To adjust the height of your Pika imaging spectrometer loosen the handle on the side of the adjustable post holder move to your desired position and then re tighten the handle Assembly of your linear scanning stage system with stabilized lighting assembly is now complete 16 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 3 2 Setup of Hyperspectral Imager 3 2 1 Connecting Pika imaging spectrometer All Pika imaging spectrometers must be interfaced to a computer for data acquisition Depending on which Pika imaging spectrometer you have this interface may utilize a USB Ethernet Firewire 800 IEEE 1394b or CameraLink interface Additionally some Pika spectrometers require external power The images below show interfacing with Firewire 800
9. INFORMATION ON FILE IMAGE ADJUSTMENTS With a few minutes of practice using the available tools you will be able to manipulate and visualize hyperspectral data quickly and efficiently 6 2 Zoom Pan Flip and Rotate Tool To zoom to a specific area of the image select the magnify tool in the toolbar and the cursor will change 46 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 Click the magnify tool in the image and the view will zoom in It is also possible to click and drag a selection within the image to zoom into the selected area wu The user may also zoom in and out using the mouse scroll wheel if available To zoom all the out and recover the original image select the original size tool and click anywhere in the image To pan the image while zoomed in select the pan tool Click and drag inside of the Image to pan Click these tool to rotate left rotate right flip vertically or flip horizontally the image 6 3 The Inspector Tool Spectral Plots eo The inspector tool allows you to see the spectrum associated with a pixel Choose the inspector from the toolbar and then click a point inside the image This will Plot the spectrum for the pixel in the spectrum plot panel List the pixel location sample and line number in the data panel ist the red R green G and blue B brightness values in the data panel 6 3 The Inspector Tool Spectral Plots 47
10. Open Spectrum Preferences Built in Scripts b Run Script Exit 7 1 2 Open Spectrum Clicking this option will open a window that allows you to browse and select a saved spectrum To see how to save a spectrum see Chapter 4 Basic Data Acquisition 7 1 3 Preferences Clicking this option opens a window that allows you to set a variety of preference options Options associated with data visualization can be found under the Plotters and workbench tabs at the top of the window The Plotters tab opens a window that allows you to control the presentation of the plots presented in the spectral plotter Additionally this window also allows you to set the parameters for exporting plot data so it can be manipulated or plotted using other software tools 61 Benchtop System Users Manual Release 4 3 The workbench tab opens a window that allows you to set the Default Image preference One of the most useful settings is RGB which generates a Red Green Blue image based on the values of three chosen hyperspectral bands which you can choose in the tool control panel Selecting the True Color button in the tool control panel produces an image that approximates the colors you would see looking at the object Adjusting the sliders allows you to generate false color images Be sure to click Update after adjusting the sliders Grayscale is another useful option which presents a grayscale black and white image based on a single hyperspectral band
11. Res New Stack Classify gt u Analyze Ratio 2 Correct 4 First Derivative Save Cube As Color 1 A S Rename Cube Ko Agricultural gt ose Cube Custom User written Plugin Datacube New Cube Analyze First Derivative This tool generates a new datacube with the first derivative of the spectral curve for each pixel A new image of the first derivative datacube is presented in the Image panel To see the spectral derivatives utilize the inspector tool eo FETE a L use one of ROI tools marquee or lasso or The first derivative curves appear in the spectral plotter Datacube New Cube Analyze Principal Component Analysis PCA A detailed discussion of PCA is beyond the scope of this document see for example Wikipedia for a discussion This tool generates a new datacube and image with the principal component values for each pixel Additionally a PCA tab appears in the tool control panel with a slider that allows you to select the number of Bands or PCA components To see the PCA component magnitudes utilize the inspector tool e or use one of the ROI tools marquee or lasso qM y or The PCA magnitude curves appear in the spectral plotter As with standard hyperspectral data classification algorithms can be performed on the new PCA datacube Correct The Correct option allows you to correct a datacube to a measured referen
12. 4 3 General purpose input e SpecBool allows the user to specify a true of false value e SpecFloat allows the user to specify a floating point value within a defined range e SpeclInt allows the user to specify an integer value within a defined range e SpecChoice allows the user to choose an item from a defined list Selecting references to cubes and spectra on the workbench e SpecCube allows the user to select one of the datacubes currently available on the workbench e SpecSpectrum allows the user to select one of the spectra currently available on the workbench Selecting portions of cubes or specta e SpecWavelength a spec that accepts a datacube as an argument and allows the user to select a wavelength from that cube e SpecBandNumber a spec that accepts a datacube as an argument and allows the user to select a single band from that cube specs have certain members that can be set to impact the way the spec appears and behaves self label The label that will appear on the graphical user interface next to the Spec s widget self units A string describing the Spec s units that will appear on the GUI if not set to None self help A string that will appear if the user hovers the mouse over the Spec 10 4 1 SpecBool class SpecBool Spec Creates boutton or a checkbox for entering a boolean def init self label defaultValue True By default a SpecBool is represented on screen by a chec
13. Again you can adjust this band using sliders in the tool control panel Be sure to click Update after moving a slider Hint Check the Auto Update box in the tool control panel so you do not have to continually click Update You may find this option is too slow for large datacubes You can also choose to open your datacubes in computer memory or your disk drive under the workbench Generally opening a datacube in memory is faster but for large datacubes or for computers with limited RAM you may find that you need to choose the disk option 7 1 4 Built in Scripts Spectronon enables you to run scripts to best suit your application A variety of scripts are built in to Spectronon and can be found in File Built in Scripts 7 1 5 Run Script You may also run Python scripts you write The Run Script opens a window that allows you to browse to your script Additional information on how to write Python scripts for your particular need is provided in Chapter 10 7 1 6 Reload Plugins When writing or changing existing plugins clicking this will make Spectronon aware of the changes 7 1 7 Exit Clicking on the Exit option closes Spectronon 7 2 Datacube Menu 7 2 1 New Image This option allows you to create a new image in the Image panel When you click on New Image a menu opens three options RGB Grayscale and Raw Camera Data Datacube New Image RGB Generates a color image based on three bands of your hyperspectral dat
14. Check binning on camera load Fig 5 9 Scanner tab in the Preferences window Note The Bin options listed above performs an average of the binned pixels This allows you to directly compare binned and un binned data 5 3 2 Workbench Preference Tab The Selection options relate to visualizing selected regions of datacubes The Floodfill options relate to the flood fill wand tool located on the menu bar which is used to select a ris a a contiguous region of spectrally similar pixels After selecting a region with either the marquee or lasso right click to reveal a pop up menu with several options Holding control while selecting ROIs allows you to append to the existing selection Metric options are either Euclidean or Spectral Angle Mapper SAM and Tolerance is a unitless parameter that determines the sensitivity of Floodfill Default Image and Open Cubes pertain to opening existing datacubes The options for Default Image are to open the image from your datacubes in a Red Green Blue RGB representation or as a single grayscale image The Open 5 3 Additional Controls a Benchtop System Users Manual Release 4 3 Warn on Unsaved Last Scan scan Settings Stat scan Scale Scan to 1 0 Update Plots Fig 5 10 Workbench tab in the Preferences window 42 Chapter 5 Advanced Data Acquisition Benchtop System Users Manual Release 4 3 Cubes options are either to memory
15. Click on the Focus F tool to toggle the camera live view off Once the imager is calibrated for both dark current and reflectance reference the imager will remain calibrated until the references are removed by the user or the machine is turned off If the integration time is changed by the user after calibration the reference signals will be adjusted accordingly To manually remove the references click Spectrometer Remove Dark Current Cube and Spectrometer Remove Response Correction Cube 4 5 Calibrate Imager 29 Benchtop System Users Manual Release 4 3 4 5 3 Adjust Aspect Ratio To scan an image type in the number of lines you would like to scan in the window just to the left of the Scan Button gt A number such as 200 is often good initially Then click on the scan button A waterfall image will appear in the Image Panel of SpectrononPro Increase or decrease the number of lines to scan as desired Once your image is scanned you can use all the visualization and analysis tools of Spectronon on your image NUMBER OF LINES TO SCAN Spectronon Pro File Datacube Image Spectrum Selection Plots Spectrometer Window Help RPL PA i OH qd p 200 Note You can stop scan by re clicking the Scan Button gt Your first image may appear distorted in one dimension To correct this distortion you must adjust the step size of your scanning system Recall that Pika imagin
16. ENVI compatible hyperspectral data can be stored in the following three formats Band Sequential In Band Sequential BSQ format each line of the data is followed by the next line in the same spectral band Hyperspectral data stored in BSQ format will end with bsq Band Interleaved by Line Data saved in Band Interleaved by Line BIL format have the first line of the first band followed by the first line of the second band followed by the first line of the third band and so forth This pattern is repeated for subsequent lines Hyperspectral data stored in BIL format will end with bil Band Interleaved by Pixel Band Interleaved by Pixel BIP data have the first pixel for all bands in sequential order followed by the second pixel for all bands and so forth for all pixels Hyperspectral data stored in BIP format will end with bip Resonon imaging spectrometers will save hyperspectral data in BSQ BIL or BIP formats and Spectronon software will open hyperspectral data in all of these formats 7 7 2 Hyperspectral data header files Header files for hyperspectral data have the same name as the BSQ BIL or BIP files followed by hdr The following discussion provides an overview of header files For additional information see for example http geol hu data online help ENVI Header Format html One can open and edit header files using WordPad for example The following list describes some but not all of header file commands Bold indicate
17. GENERATE THE IMAGE i F 460 0 al 404 6985 699 S605 934 4625 Blue Mode Wavelength To generate false color images use the sliders or arrows to change the wavelength bands used to create the RGB Image This tool is often useful when trying to visualize specific spectral features associated with an object in your image If Auto Update is not selected click Update to generate the new image The Mode menu allows you to identify the band by wavelength typically the most useful or by band number As an example of how false color images can reveal interesting features move the red slider to approximately 593 nm and the green slider to approximately 516 nm then click Update This generates a new false colored image shown below that reveals there are actually two kinds of red candy and suggests there are two kinds of yellow candy each candy type is positioned in the shape of an P In Chapter 3 a classification technique will show this more clearly 54 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 Spectronon Pro Datacube mage Spectrum Selection Plots Spectrometer Window Help Sh ta Took mene Orange Yellow Pw0S RGs Fees rc Tiaa Specktregranm X Y Cubes Orange Yellow PAU S E GB 300 Sperra 2700 M Calbration 1800 xol 400 500 600 700 BOO 900 1000 Condos update Auto Update p 1 RGB Contrast Cube C Documents and S
18. Hf Searches datacubes bip bil bsq xv Expand the directory tree on the left side of the window to find the datacubes you wish to process Use the buttons in the middle of the screen to select the desired datacubes Datacubes that will be processed appear on the right side of the window In the figure above 3 datacubes are selected for processing When you are happy with your selection click Ok A new window appears to define the plugins you wish to run 115 Benchtop System Users Manual Release 4 3 Batch Processor Select the plugins to apply Cube plugin sequence Airborne Radiance Conversion Normalized Difference Vegetative Index NDVI Airborne Radiance Conversion Airborne Reflectivity Conversion Correct From Cube Correct From Measured Reference Correct From Spectrally Flat Reference Radiance Conversion Reflectance Conversion Color 5 Agricultural i Anthocyanin Reflectance Index 1 Anthocyanin Reflectance Index 2 SS REMOVE Renders produced from final cube of sequence Atmospherically Resistant Vegetative Index Carotenoid Reflectance Index 1 Carotenoid Reflectance Index 2 Enhanced Vegetative Index Modified Chlorophyll Absorption Reflectance Index Modified Red Edge Normalized Vegetation Index Modified Red Edge Simple Ratio Index Normalized Difference Vegetative Index NDVI Photochemical Reflectance Index Render ExportFormat Pl
19. I a 1t M6 gt gg ng A eMachines 7 i Do NOT install the lens onto the imager if the barrel is screwed in too far as shown below 3 2 Setup of Hyperspectral Imager 19 Benchtop System Users Manual Release 4 3 While holding and squeezing the chrome collar thread the objective lens onto the imager The objective lens is now installed 3 2 4 Setting Objective Lens Aperture Next set the f stop f of the objective lens There is a range of acceptable options and the end user will need to choose the appropriate value based on his or her application Large f s will provide large depths of field but limit the amount of light collected whereas small f s will have small depths of field but will exhibit better light gathering 20 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 Screw ES m ames ALIGNMEN To set the f on Schneider lenses for example loosen the setscrew on the objective lens itself This will allow you to rotate the f collar on the objective lens which can be read by noting the location of a small white alignment line Align the white line with your desired f then retighten the setscrew Acceptable f s Pika II 3 and larger Pika XC 2 4 and larger Pika NIR 1 8 and larger 3 2 5 Installing SpectrononPro Software if necessary Your hyperspectral imaging system will be delivered with SpectrononPro already installed If however you need to reinstal
20. RGB Class TriBand Return an RGB 34ayer image from 3 given bands of a cube Tools Controls Camera Stage Filter Tool Contrast Class Stretch J Auto Update Stretch the brightness values between two give percentage amounts RGB Contrast Presets True Color Color Infrared Red J 374 525 661 295 948 065 87583754 127 Green J line 91 374 525 661 295 948 065 R 1521 cle 4634237 G 808 B 529 Blue 374 525 661 295 948 065 Mode Wavelength v INFORMATION DATA PANEL TOOLS PANEL PANEL Once the software has started make sure that the imager and stage controls if used are enabled The imager and stage tools will be greyed out if not enabled as shown below Spectronon Pro File Datacube Image Spectrum Selection Plots Spectrometer _Wjpdoismebh ENABLED Spectronon Pro File Datacube Image Spectrum Selection Plots e E NOT ENABLED 24 Chapter 4 Basic Data Acquisition Benchtop System Users Manual Release 4 3 You can get the latest version of SpectrononPro by clicking on Help Check For Updates This won t download the latest version but will give an alert if there is a newer version 4 2 Camera Controls Exposure parameters can be controlled by clicking on the Camera tab in the Tools Panel Tools Controls Camera Stag rn Camera Settings AE 53 17 Framerate hz 0 04 31 07 62 11 AP 7 809 0 0 01 50 005 100 0 ME
21. a number of options As will be seen later the ROI tool is often used during one of the first steps in classifying different objects within a hyperspectral image To select a Region of Interest ROD select either the marquee lasso or flood fill wand tool from the menu bar Click and drag a rectangle of interest with the marquee tool or click and drag any closed shape with the asso The floodfill tool can be used to select a contiguous region of spectrally similar pixels After selecting an area right click to reveal a pop up menu with several options Holding control while selecting ROIs allows you to append to the existing selection Hint The selection menu is also available in the main menu The floodfill tool shows pixels that are spectrally similar to the chosen pixel Spectral similarity is assessed with either Euclidean distance or Spectral Angle Mapper SAM along with a tolerance value The user can set these options by 48 Chapter 6 Basic Data Analysis Spectronon Pra Benchtop System Users Manual Release 4 3 Fie Dasube Lm ainm Selain Plo Spectrometer Windew Info Resource Tree Py cubes eg orange Grange Yellow FACIS RI Seer bes M Calbration Info Cube Ciipocuments an Fiber Tad Certract Clas SEpetch bres breghbniess vas bet uen han pre percentage amounts Data 512 line f56 3479 3054 B 102
22. background fluorescence of the substrate Classify The Classify option allows you to generate classification maps of different objects within your hyperspectral data More information on hyperspectral classification can be found in Chapter 8 Spectronon Pro File Image Spectrum Selection Plots Spectrometer Window Help New Image MEZ estera 500 Info New Cube Utilities gt Res New Stack Classify 4 Euclidian Distance T oe Analyze 4 Spectral Angle Mapper Correct 4 Spectral Unmix save Cube As b 3 Rename Cube ini X Close Cube Agricultural Custom User written Plugin Some of the classification algorithms SAM Euclidean Distance Spectral Unmix use Spectrum objects as inputs while the statistics based classifiers Logistical Regression Discrimient Analysis Support Vector require datacubes 64 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 of the classification target These are made selecting the ROI with the lasso or margee tool then selecting Create Cube from Selection When you select one of the classification options e g SAM a new cube will be generated in the datatree panel In many plugins you will need to specify how many Layers you wish to classify 1 e how many materials you wish to classify you will then need to select a Spectrum or Cube for each of these Layers in a pull down menu The Spectra and Cubes avail
23. imaging spectrometers are line scan imagers which means they collect data one line at a time To assemble a complete two dimensional image multiple lines are imaged as the object or imager is translated The multiple line images are then assembled line by line to form a complete image as indicated below A scanning system is often needed to use a Resonon imaging spectrometer Multiple lines are assembled to create a two dimensional image Em Leaf Scan Direction To obtain hyperspectral data the signal from each pixel is dispersed or diffracted into its spectral components much like passing the light from each pixel through a prism This process occurs for every pixel in the line red squares in the image below The dispersed signals from all the pixels are imaged onto a focal plane array One of the benefits of this approach is that all the spectral color information is collected at the same time for each pixel The result is a detailed spectral curve for every pixel in the image 4 Chapter 1 Introduction to Hyperspectral Imaging Benchtop System Users Manual Release 4 3 Benchtop System Users Manual Release 4 3 6 Chapter 1 Introduction to Hyperspectral Imaging TWO SYSTEM OVERVIEW Resonon s benchtop hyperspectral imaging system is comprised of a Pika hyperspectral imaging camera linear trans lation stage mounting tower lighting assembly and software control system The positions of the i
24. init self label datacube defaultValue None A SpecWavelength accepts a datacube as an argument and allows the user to select a wavelength that is present in that cube It s uder interface is a slider with range equal to the wavelength range in the passed in datacube Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecWavelength class MyPlugin CubePlugin 106 Chapter 10 Custom Data Analysis Writing Plugins Benchtop System Users Manual Release 4 3 An example plugin label My Plugin userLevel 1 def setup self self my wavelength SpecWavelength label Wavelength datacube self datacube defaultValue None self my_wavelength help my mouse hover help def action self a float wavelength self my_wavelength value a band at a user selected wavelength band self datacube getBandAtWavelength self my wavelength value 10 4 8 SpecBandNumber class SpecBandNumber SpecInt def init self label datacube defaultValue 0 A SpecBandNumber allows the user to select one of the bands present in a datacube by index Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecBandNumber class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my band num SpecBandNumber label Band datacube self datacube defaultValue 0 self my_band
25. or disk The memory option is faster but may not work well for computers with limited memory When checked the Warn on Unsaved option will provide you with a warning message that you have not saved your last scan when closing the software Similarly when clicked the Last Scan option will provide you with a warning message that your previous scan will be overwritten by your next scan if you have not already saved it The Default Scan Image option allows you to set whether your scanned image is displayed as a Red Green Blue RGB image or as a grayscale image This setting does NOT impact the hyperspectral data itself it only changes how the image of the data is presented in the image panel The Recording option allows you to record your scans into computer memory or directly to disk Recording to memory is faster and preferred unless your data acquisition computer does not have sufficient memory for your scans Be sure to click on the Set button if you change this setting The Scale Scan to 1 0 option allows you to record reflectance as a floating point number with the brightness scaled to 1 for your reference sample This saves your data as a standard reflectance value IF you use a reference sample with a reflectivity of 1 Spectralon is a good material for this and Teflon is a less expensive and reasonable substitute reference material although less accurate If you use the Scale Scan to 1 0 option each channel of data will require 4 bytes
26. shown Maximize the sharpness of the lines a ea 3 m hd X PLOT REVEALS CROSS SECTION WITH INSPECTOR TOOL e TYPICAL WELL FOCUSED IMAGE Hint Clicking the Inspector Tool e in the Image Panel and then selecting X tab in the Plots window will reveal a cross section plot of your image Viewing this plot allows you to graphically see the sharpness of your focusing You can zoom in on the X axis by clicking the Zoom tool Es and then clicking on the X axis Once you have completed focusing re tighten the lock to the focus adjustment Then click on the Focus tool again to toggle the camera live view off Hint See our You Tube video on focusing at http www youtube com resonon 4 5 Calibrate Imager The following discussion describes how to set up your system to scan for reflectance scaled to a reference panel More advanced options are described in the Appendix 28 Chapter 4 Basic Data Acquisition Benchtop System Users Manual Release 4 3 4 5 1 Remove dark current SpectrononPro makes it easy to remove the average dark current noise from your scans Begin by clicking on the Dark Current button D on the SpectrononPro toolbar You will be instructed to block all light entering your Pika imaging spectrometer by blocking the objective lens Please block all light te lens with the lens cap and click Ok or Cancel to skip Dark Current Recording Once you have t
27. title Post a message a dialog box postScrolledMessage self message title Post a message in a dialog box that can be scrolled and text can be selected Good for long messages postQuestion self message title Proceed post message returns True if user selects and False if Cancel params message the message to show title the dialog box title requestOpenFilename self wildcard message Enter Filename To Open multiple False Requests a filename s from the user Returns the filename as a string or a list of strings if multiple True Returns None if the user selects Cancel Chapter 10 Custom Data Analysis Writing Plugins Benchtop System Users Manual Release 4 3 def regquestDirectory self message Please give a directory path suggest wee Reguests a directory form the user Returns the selected directory path as a string Returns None if the user selects Cancel 10 6 Using Datacubes Plugins get a reference to their relevant datacube as self datacube It is possible to get references to other datacubes directly from the workbench Datacubes contain a data array in numpy format as well as a collection of metadata Some important methods you can call are def getSpectrumArray self sample line asFloat False return a 1D array at location sample line params sampl
28. written Plugin 68 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 To transform the hyperspectral data into CIE Colorspace choose the pathway Datacube New Cube Color CIE Colorspace Conversion This will generate a new datacube with CIE colorspace values for each pixel as well as a new CIE Colors pace Conversion tab in the tool control panel In this window you are provided with standard options for the Observer and Illumination To see the Color space Values position the inspector tool eo on a pixel and its values will appear in the Data panel Note A warning will appear if your datacube does not have all the spectral values needed to perform a CIE color space conversion 380 nm to 720 nm Click and Spectronon will utilize the bands available to make the appropriate conversion based on the data available You can determine how close the pixels are to desired Color space values by using the AE option To do this you first have to perform the CIE Color space conversion as described above then follow the path Datacube New Cube Color Delta E This will generate a new image in the Image panel and a new tab in the tool control panel called Delta E A slider appears in the tool control panel that allows you to select how many AE values you wish to compute Below this sliders appear that allow you to set your target LAB values After setting your target LAB values clic
29. you want to close it Clicking OK will close the datacube Clicking Cancel will cancel the command so you can continue working with the datacube or save it 7 3 Image Menu The Image menu provides tools for saving or closing images generated from your datacube Image Spectrum _ Selection Export Image Export Image As KML Close Image 7 3 1 Export Image Clicking on Export Image will open a window that allows you to name and save the image file at a location of your choosing Additionally you can choose your preferred image format Options include TIFF PNG BMP JPG and GIF 7 3 2 Export Image as KML This allows the user to export an image in Keyhole Markup Language KML format 7 3 Image Menu 71 Benchtop System Users Manual Release 4 3 7 3 3 Close Image This tool will close the image shown in the current Image panel 7 4 Spectrum Menu The Spectrum menu provides tools for manipulating individual spectrum 7 4 1 Save Spectrum The Save Spectrum tool will save the spectrum selected in the Resource Tree in its current configuration If the spectrum has not been previously saved a window will open that allows you to rename the spectrum and browse to save it in the file of your choosing Once saved the new name will appear in the spectral plot panel The color selected for the spectrum will be saved Spectrum Selection Plots Spectrom Save Spectrum Save Spectrum As Export Spectrum As Text Fi
30. 0 Gain dB 0 17 34 Frame Rate is equal to the number of images acquired each second and limits the maximum exposure time Max Exposure Time 1 0 Frame Rate Integration Time also known as Exposure Time is the duration of data acquisition for each individual line image Gain is a factor which increases the signal but at the expense of signal to noise ratio Try to keep gain as low as possible preferrably zero unless absolutely necessary Note The Pika NIR camera does not have a gain control tool 4 3 Stage Controls You can move the stage manually by clicking on the Jog Stage buttons located on the tool bar of SpectrononPro The buttons will move the stage incrementally in either direction Use the buttons to center the stage underneath the Pika imaging spectrometer 4 2 Camera Controls 25 Benchtop System Users Manual Release 4 3 JOG STAGE Spectronon Pro BUTTONS File Datacube Image Spectrum Selection Plots Spectrometer Window D P 500 p gt fps The stage can be further controlled by clicking on the Stage tab in the Tools Panel Tools Controls Camera Stage Linear Speed Units Rotation C Motor Stepping Mode Scanning Speed J cms 0 07938 7 97719 15 875 Go Home After v Homing Speed cms 15 875 Jog Speed J cms 0 07938 7 97719 15 875 Speed Units is setting used for different types of stages Linear is used for the st
31. 0 dimensions for 100 band hyperspectral imagers In general additional dimensions provide additional opportunities for points from different objects to be distinct in the color space and thus they are easier to classify in practice but otherwise the number of dimensions need not concern us now Admittedly it is difficult to visualize a 100 dimension space but the mathematical transition is often straight forward Consequently one can invent techniques that work in two dimensions and they can generally be applied to 100 dimensions Consider the following example of the well known hyperspectral classification algorithm known as Spectral Angle Mapper SAM Note that the groups of training pixels from the three different objects in our 2 dimensional color plot are located at different angles relative to the horizontal axis The SAM technique utilizes this property to classify ALL pixels Consider a reference vector at the center of the group of purple training pixels This vector is shown as a large orange vector below Similarly we can envision vectors from the origin to each pixel in the plot only three are shown to reduce clutter 80 Chapter 8 Advanced Data Analysis 2 Hyperspectral Classification Benchtop System Users Manual Release 4 3 Red REFERENCE VECTOR FOR THE PARALLELOGRAM ORANGE AND FOR THREE OTHER PIXELS Blue One can see that the angle between the reference orange vector is small for the purple pixels and
32. 1 964 485 876 431 783 496 7 501 612 Photochemical Reflectance Index Plant Senescence Reflectance Index Red Edge Normalized Difference Vegetation Index Simple Ratio Index Structure Insensitive Pigment Index Transformed Chlorophyll Absorption Reflectance Index Vogelmann Red Edge Index 1 Vogelmann Red Edge Index 2 Vogelmann Red Edge Index 3 505 524 511 435 515 5483 571 765 Water Band Index Data Statius Choosing HVI from this menu will generate a grey scale image 92 Chapter 9 Advanced Data Analysis 3 Hyperspectral Vegetation Indices Benchtop System Users Manual Release 4 3 File Datacube Image Spectrum Selection Plots Window Help amp te VNIR arborne datacube 1 ARI2 Greyscale Info Resource Tree fJ Cube 3 VNIR airborne datacube 1 CJ VNIR airborne datacube 1 RGB e ARR ARD Greyscale EJ wcaRI MCARI Greyscale Spectra Calibration Resource Info Cube MCARI Render Tool Greyscale Class SingleBand Return a single band rendering from one of three selection techniques Filter Tool Contrast Class Stretch Stretch the brightness values between Use Contrast two give percentage amounts y Presets 2 stretch Linear stretch Adjustments Low Cut Percentage High Cut Percentage Inverse Individual Bands Renamed cube VNIR airborne datacube 1 MCRI to MCART The value
33. 3 Hyperspectral Vegetation Indices Benchtop System Users Manual Release 4 3 File Datacube Image Spectrum Selection Plots Window Help Info 4 VNIR airborne datacube 1 RGE From5Selection Resource Tree T Cubes FJ VNIR airborne datacube 1 O airborne datacube 1 RGB Fromselection VMIR airborne dataci E FromSelection VNIR airborne da Fromselection VNIR 4 H C FromSelection Mew Image Greyscale A Spectra New Cube RGB mm VNR airborne datacube Hew Stack Scalar to color map Wy i aliblwratinen Lave Cube Resource Info Save Cube As FromSelection VNIR airborne data Rename Cube bands 1 Close Cube ines 79 interleave bip samples 1483 refiectance scale factor 1 0 data type 4 ratabon 0 0 0 1 1 0 1 1 band names Modified Chiorophyl Absorption Reflectance Index map info 1 000 1 000 435155 0 5063932 67 0 2000 0 2000 12 North WGS 84 units meters rotabon 0 0007 history Acbmar Crop gt Below is a spectral colormap of the MCARI index for the example datacube 9 2 Generating HVI maps in Spectronon 95 Benchtop System Users Manual Release 4 3 e Cubes C airborne datacube 1 0 airborne datacube 1 RG ARD Greyscale ARD Scalar to color map E mcan MCARI Greyscale 0 MCARI Scalar to color m Snectra gt m j sel 0 0 25 0 5 0 75 Wavelength nm
34. 4095 1023 1 If you divide the data by this factor the data will scale from 0 to 1 typically for reflectance data byte order 0 byte order is the ordering of byte significance http en wikipedia org wiki Endianness Data for Spectronon are LSF or Little Endian which means the least significant bytes come first Spectronon will not properly open BSF or Big Endian files header offset 0 This is the number of bytes at the beginning of the binary file that should be ignored wavelengths List of the spectral band centers in units of nm rotation 4 4 This is a default orientation of the view of a datacube in Spectronon This is Spectronon only header option label name The label is used by Spectronon to give a human readable alternative to the file name of a file If omitted Spectronon will use the file name timestamp day amp time of datacube If available this is when the datacube was recorded The timestamp is often found on airborne datacubes and datacubes where a script entered this information explicitly into the header The following are for spec hdr files only headers for spectrum Spectrum files are ENVI compatible because they are 1x1 datacubes but Spectronon adds in these data for its own use original cube file name of original cube This is a Spectronon only header item for recording the history of a generated spectrum In the case of a derived da
35. 7 5 3 Mean Spectrum This tool will calculate and plot the Mean Spectrum in the spectral plotter for the pixels selected in your ROI A plot color will be automatically assigned to the plot and the spectrum will be listed in the Resource Tree 7 5 4 Create Cube From Selection This tool will generate and display a new datacube from the selected ROI An image of your new datacube will appear in the Image panel and the new datacube will be listed in the Resource Tree This tool allows you to easily crop your datacube into smaller pieces when you wish to concentrate on small regions of your original image or to create complete new cubes of regions of interest for quantitative analysis If the selected ROI is not rectangular spectronon will create a rectangular representation of the pixels within the selected ROI The broadcast of non rectangular ROI to a rectangular shape may result in loss of spatial information but all spectral information will be retained 7 5 5 Mean and Standard Deviation This tool is much like the Mean Spectrum option except that it will also plot curves in the spectral plotter corre sponding to the mean spectrum plus and minus the standard deviation of the brightness values within the ROI This option enables easy visualization of the variability in your spectral data as a function of wavelength 7 5 6 Mean First Derivative This tool plots the first derivative of the mean spectral curve within the ROI as a function of wavel
36. A ea T2 Datacube WICH c o 05 aasma d Be o oe dE 49 eR Re do Image Ment ase Ree x9 42 666 HS 9399363 43 T Nien 2o vxo E URREE SRVEPBeLeeGUERSG Ao SEEC Men 6 s e ooo SR RR od wor srete mee RP Re X HE ES 7 7 Datacubesand Header Files Advanced Data Analysis 2 Hyperspectral Classification CONTENTS 9 10 11 5 General Approach 8 2 Spectral Angle Mapper SAM Classification Advanced Data Analysis 3 Hyperspectral Vegetation Indices aa ASR OR a aa Ee OE SO 92 Generating HVI maps in Spectronon Custom Data Analysis Writing Plugins IOI Getma rua ea ma reasda a e OS e a e a ee S a 10 2 Specronon Plugim OyervieW gt 6244565445 68 be GEG EG Ee kad 10 3 The attributes available to each plugin type DEE cce E wes OR S A CORE S AEPSSSRSSCELSEXS4q Bes eee 105 Using the WOPEDBBO s e s esras 9 6a lOO tune DAPIGUDES sa ss xus 33 OPERAS RS CE EOS aa a a ud 10 7 Example PIMOS s2 646 4 4 AFG SSS dia 3X9 hni AGRO 908 X 9o 33799 EEE HES Batch Processing 12 Focusing amp Calibration Sheets 13 14 124 Small Focusing Sc kha ct eee eRe Ee EH R POE POR hob mono x 4 12 2 Large Focusing Sheet cg Gok
37. BandCount Most commonly you will pass your plugin s datacube member to SpecCube s constructor as the datacube argument but any datacube object is a valid choice You should always pass the plugin s self wb workbench member to SpecCub as the wb argument The value of a SpecCube is an index to a cube within the workbench To get the cube object itself you must call the wb tree getCube method Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecCube class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my cube SpecCube label Select a Datacube datacube self datacube wb self wb requireMatchedLineCount False requireMatchedSampleCount False list only those cubes with the same number of bands as the cube we operate on requireMatchedBandCount True requireBandCount None defaultValue None self my cube help my mouse hover help def action self primary cube self datacube secondary cube self wb tree getCube self my cube value operate on the datacubes 10 4 6 SpecSpectrum class SpecSpectrum Spec Displays a ComboBox for selecting one of the loaded spectra def init self label datacube 10 4 Specs 105 Benchtop System Users Manual Release 4 3 wb requireMatchedWavelengths False requireMatchedBandCount False defaultValue None SpecSpectrum
38. Benchtop System Users Manual Release 4 3 Resonon Inc December 07 2015 Introduction to Hyperspectral Imaging System Overview System Assembly 3 1 Assembly of Stage Tower and Lighting 3 2 1 Basic Data Acquisition Al Start The System 4 6x soo e ck omo bee o3 RR ORO 3 Rom o3 RR SR X S 4 2 Camera ae Gom o9 X 4 3 Stage Controls 44 Focus Objective Lens 4 5 Calibrate Imager 4 6 Scanning and Saving Datacubes Advanced Data Acquisition ol Camera 4446250 db x RR EG oo HEHE ee nene 2 2 Stage ContT ls amp sonus re sahid REE ERE ERED 2 9 Additonal Controls s 29 9 aoo ox rka YE wee EET EGG EG Basic Data Analysis Ol Spectromon Tools 62 Zoom Pan Flip and Rotate Tool 6 3 The Inspector Tool Spectral Plots 64 Region Of Interest Tools 6 5 mage lt gt e45eebe Ro ea 3 RB X om FRESE eS OO uu wu abs em 33km o ED o BOR Wu xu X XS 6 7 Saving 1 1 lt 6 5 More Information gt Advanced Data Analysis 1 General Tl Xeon koc chew ox do ocho m 3 EMG HOD dom DE
39. File Datacube Image Spectrum Selection Plots Window Oo BE D C te Info VNIR airborne datacube 1 RGB Resource Tree PP Cubes a WIR git at rj VNI New Image ZA A Spectra Cube k Utilities Em VIR New Stack Classify k M Calibration 7 T Save Cube RM Save Cube Sate Rename Cube CON i A T Close Cube Agricultural d Anthocyanin Reflectance Index 1 Clustering k Anthocyanin Reflectance Index 2 B Resource Info Band Ratio Example Atmospherically Resistant Vegetative Index WHER sirbome datacube 1 Custom User written Plugin Carotenoid Reflectance Index 1 nma 1994 4 k Carotenoid Reflectance Index 2 samples 1 lines 2071 Enhanced Vegetative Index Modified Chlorophyll Absorption Reflectance Index ee Modified Red Edge Normalized Vegetation Index Modified Red Edge Simple Ratio Index Mormalized Difference Vegetative Index NDVI byte order 0 Zone 12 H LR Easting km 495 15650 435 85240 0 20 m B bt northing km S063 22507 5063 936567 0 20 m dem automatic fram scc dem data base gps c users resonon desktap sample data for reflectivity conversion Xubeil20100914 20h3 amp m hs ibil tet fov 17 000 deg fipped Ines 0 to 1193 no rotation rph biases deg 0 000 0 000 0 000 wavelength 408 284 413 196 418 108 323 02 427 932 432 844 337 756 442 668 447 58 452 492 457 904 M 462 316 457 228 472 14 477 052 48
40. O Orange PWw s RGE 3 Orange Yellow FW Ors Gneysc Spectra M Calbestion DO 500 600 700 BOD 900 1000 E Info Cubs and P Aho E Settings Rand Desktop Maie loan cte gp ge Yeka PAG 4 Band Wavelength ww 305 700 284 Bender Took Greyscale Class return single band rendering from one of thee selection techniques Fiker Tool Contrast Class Stretch Mode Wavelength Sbrebch thee brightness values etait Eo give parien iig amounts Data Sample 2563 ire 255 006000000 000000000 900000 006 90000000 Yau 2504 As with the RGB images a vertical line in the Spectral Plot shows the band you have chosen Note that even though the image is from a single band the Inspector and ROI Tools will continue to plot and operate on all wavelengths Hint With Auto Update selected in the tool control panel you can quickly scroll through single band images Warning For large datacubes or slow computers the Auto Update refresh rate may be slow 6 6 Plot Panel The plot panel allows you to visualize hyperspectral data graphically This has already been seen with the use of the Inspector and ROI tools but here we explore the plot panel in more detail The plot panel has three tabs Spectrogram X and Y These three tabs provide you with plots along the three axes of a datacube using the Inspector Tool eo as shown below 6 6 Plot Panel 57 B
41. ROI region for the plot 7 4 10 Copy Spectrum as Text This tool allows you to easily transfer the spectral data to another program such as Excel or Notepad Select the spectrum you wish to copy in the Resource Tree and then click on Copy Spectrum as Text This copies the spectral data onto your clipboard You can now paste the data into a program of your choosing Hint Positioning your mouse on a spectrum in the Resource Tree and right clicking will open the Spectrum menu items described above 7 4 11 Crop Spectrum To use this tool first select a spectrum then select a region in the spectra window and then click Crop Spectrum The result will be a graph of the spectra in the region chosen Note this tool can be accessed by right clicking the mouse after selecting a spectra 7 5 Selection Menu The Selection menu items provide a number of options for use with the selection tools Or For all of the following options you must first select a Region of Interest ROI 7 5 Selection Menu 73 Benchtop System Users Manual Release 4 3 Select All Select None Mean Spectrum Create Cube From Selection Mean and Standard Deviation Mean First Derivative Send To Clipboard Note The Selection menu can be accessed from the main menu or by right clicking in the image after creating an ROI 7 5 1 Select All This tool selects the entire dataset 7 5 2 Select None This tool clears the current selection
42. SYSTEM ASSEMBLY 3 1 Assembly of Stage Tower and Lighting Begin system assembly by removing all items from the shipping package Confirm that the items shown in the images below If you are missing components please contact Resonon LIGHTING ASSEMBLY REGULATED POWER SUPPLY T NUTS AND BOLTS TOWER WALL WART POWER CONNECTOR ADJUSTABLE POST HOLDER USB CABLE BASEPLATE AND LINEAR STAGE Install lighting onto the tower Begin by sliding the T nuts into the slots on the tower Slide in one T nut into each slot on the same side as the adjustable post holder Benchtop System Users Manual Release 4 3 SLIDE T NUTS INTO THE GROOVES OF THE TOWER Position the T nuts side by side at the location you wish to attach your lights Note You may easily adjust the lighting position later POSITION T NUTS SIDE BY SIDE AT THE DESIRED LOCATION FOR YOUR LIGHTS Place the lighting assembly on the post such that the mounting holes line up with the T nuts 10 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 POSITION LIGHTING ASSEMBLY OVER THE NUTS Attach the lighting assembly to the T nuts using the supplied bolts TIGHTEN BOLTS WITH F AN ALLEN WRENCH TO TO MOVE THE SECURE THE LIGHTING ASSEMBLY 7 LOOSEN BOLTS SLIGHTLY AND SLIDE THE ASSEMBLY TO YOUR DESIRED POSITION Next attach the post to the baseplate Begin by removing the four bolt
43. Use the focusing sheets to focus the objective lens and use the aspect ratio calibration sheet to set the stage speed and imager framerate See the chapter on Basic Data Acquisition for details 12 1 Small Focusing Sheet Focusing Sheet 119 Benchtop System Users Manual Release 4 3 120 Chapter 12 Focusing amp Calibration Sheets Benchtop System Users Manual Release 4 Focusing Sheet Benchtop System Users Manual Release 4 3 12 3 Aspect Ratio Calibration Sheet Aspect Ratio Calibration Sheet scan direction increase stage speed or calibrated 1 1 ratio decrease stage speed or decrease frame rate increase frame rate 122 Chapter 12 Focusing amp Calibration Sheets Benchtop System Users Manual Release 4 3 12 3 Aspect Ratio Calibration Sheet 123 Benchtop System Users Manual Release 4 3 124 Chapter 12 Focusing amp Calibration Sheets THIRTEEN CONTACT US FOR PRODUCT SUPPORT Phone 406 586 3356 Website http www resonon com Email productsupport resonon com Also see our product support webpage located at http www resonon com customer_support html 125 Benchtop System Users Manual Release 4 3 126 Chapter 13 Contact us for product support FOURTEEN COPYRIGHT NOTICE Copyright March 2014 Resonon Inc rights reserved No part of this publication may be repr
44. _num help my mouse hover help def action self band self datacube getBand self my_band_num value 10 5 Using the Workbench Plugins automatically receive a reference to the Spectronon workbench self wb The workbench gives you a lot of freedom including the freedom to break things so use this reference wisely Most often you will use the workbench to get references to other datacubes or spectra on the resource tree add additional items to the resource tree or prompt the user for a filename or other information SourceIDs in the below methods refer to IDs returned from SpecCubes or SpecSpectra Some important methods you can call are 10 5 Using the Workbench 107 Benchtop System Users Manual Release 4 3 def def def def def def def def def 108 getCube self sourceID zNone Returns the requested datacube sourceID is None returns the current cube getRendering self renderID None Returns the requested rendering sourceID is None returns the current rendering getSpectrum self sourceID None W ni Returns the requested spectrum sourceID is None returns the current spectrum addCube self datacube name None render True Adds datacube to the workbench addSpectrum self spectrumObject name None Adds a spectrum to the workbench postMessage self message
45. a These bands can be chosen and adjusted using sliders in the tool control panel 62 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 Datacube New Image Grayscale This produces a single band image from a single hyperspectral band that can be chosen using a slider in the tool control panel Spectronon Pro File Image Spectrum Selection Plots Speci New Image 4 Greyscale Info New Cube RGB R New Stack 3 NM Save Cube Save Cube As Rename Cube X Close Cube 7 2 2 New Cube Utilities The Utilities option allows you to generate a new modified datacube from the currently open datacube Spectronon Pro File Image Spectrum Selection Plots Spectrometer Window Help Gees EG Gd 500 New Image Info New Cube gt Utilities 4 Bin Cube Res New Stack Classify gt Convert To BIP T gt Crop di Save Cube As orrect gt Scale To One Utility AS Color gt Subtract Spectrum X C ince Cube Agricultural LL Custom User written Plugin Datacube New Cube Utilities Bin Cube Generates a new cube by binning spectral and or spatial channels which is often useful to either reduce the size of the datacube or improve the signal to noise ratio This choice creates a new image in the Images Window and also generates a Bin Cube tab in the tool control panel with 3 sliders Th
46. a ae eR ORS OEE ES eS EE 12 3 Aspect Ratio Calibration Sheet lt is s OX RO ESSE HE OE OR d HS Contact us for product support Copyright Notice 115 119 119 121 122 125 127 Benchtop System Users Manual Release 4 3 CONTENTS 1 Benchtop System Users Manual Release 4 3 2 CONTENTS INTRODUCTION TO HYPERSPECTRAL IMAGING Hyperspectral Imaging or imaging spectroscopy refers to the creation of a digital image containing very high spectral color resolution Each spatial point pixel in a hyperspectral image represents a continuous curve of incoming light intensity versus wavelength For example red green and blue arrows below show the spectra for three pixels of an image of a leaf in the image below The data can also be interpreted as a stack of images with each layer in the stack representing the scene at a different wavelength this stack of two dimensional images is referred to as a datacube zu roo Wavelength nmi The benefit of the additional information provided by hyperspectral imaging is that it generally allows you to tell the difference between materials better than you can with traditional color images This capability can be applied to a wide range of current and emerging applications in remote sensing sorting quality control research and development and more Benchtop System Users Manual Release 4 3 Resonon
47. able in the pull down menu can be generated either by using one of the ROI tools marquee or lasso or or by loading a previously saved spectrum or cube Note a detailed discussion of hyperspectral data classification is beyond the scope of this document Spectronon provides some of the more commonly used algorithms For more advanced algorithm capability please see other packages such as ENVI Additionally custom algorithms can be utilized with Spectronon using the scripting capa bility described in Chapter 9 The classification options provided are Datacube New Cube Classify Euclidian Distance This is a commonly used classification algorithm for hyperspectral data that is more sensitive to pixel brightness than SAM Datacube New Cube Classify Spectral Angle Mapper SAM The SAM classification routine is described in detail in Chapter 6 Datacube New Cube Classify Spectral Unmix Spectral unmixing deconvolves the signal in each pixel into a linear combination of known spectra Analyze The Analyze option allows you to perform useful operations on hyperspectral data that will generate new datacubes based on applying analytical functions to your datacube 7 2 Datacube Menu gg Benchtop System Users Manual Release 4 3 Spectronon Pro File Image Spectrum Selection Plots Spectrometer Window Help 9 gt LESE 5 T edic 500 Info New Cube Utilities gt
48. al you may increase the Gain manually The Auto Expose button in the toolbar will always set the gain to zero Note The Pika NIR camera does not have a Gain control tool Hint When adjusting the camera settings you may observe the effects of your adjustments in live view To do this click on the focus button Note If you observe broken or torn images when recording data at high speeds you may be able to eliminate this problem by collecting data at slower speeds These errors may occur due to a number of shortcomings within the data acquisition system including limitations in disk write speed chipset motherboard bus speed and CPU speed 5 1 2 Slope and Intercept Calibration Values As discussed in Chapter 3 System Assembly the slope and intercept values should be input when you first set up your hyperspectral imager The values are provided on a paper calibration sheet with your imager Contact Resonon support if you have lost your slope and intercept values Once these values are set they should not be changed 5 1 3 Auto Exposure Parameters Spectronon features a user configurable auto exposure algorithm that can be controlled in the camera settings tab Two parameters are available Target Brightness and Target Pixel During the auto exposure routine Spectronon calculates the histogram of pixel brightness of each frame It then iteratively adjusts the integration time such that the histogram percentile represented by t
49. andAverageExample RenderPlugin returns a greyscale image of the average across all bands label Band Average Example userLevel 1 10 7 Example Plugins 113 Benchtop System Users Manual Release 4 3 def action self datacube self datacube create an numpy array of the same spatial dimensions of the datacube out numpy zeros datacube getLineCount datacube getSampleCount dtype numpy float32 for band in range datacube getBandCount for each band add it to the accumulator out datacube getBand band normalize the results by the number of bands out out datacube getBandCount return out 10 7 3 Filter Plugins Filter Plugins operate on a render to enhance or otherwise filter its appearance It is currently not possible for a user to write a Filter Plugin 10 7 4 Select Plugins Select Plugins operate on a Region of Interest created in Spectronon with the Lasso Wand or Marquee tools An example is shown below that finds the spectral and spatial average value of the pixels inside of the ROI from spectronon workbench plugin import SelectPlugin class AverageROI Example SelectPlugin Calculate a single mean across all bands selected region label ROI Example userLevel 1 def action self get the raw numpy data from the datacube a BIP is indexed as lines samples bands cube_array self datacube getArray asBIP True Use a numpy slice to get
50. andard linear translation stage that is installed on most benchtop systems Rotation is used for a tripod mounted rotational scanning stage typically used in outdoor applications Motor displays speed in motor pulses per second Stepping Mode controls the way the stage moves in relation to the imager When the Stepping Mode box is not checked the stage and imager both run continuously during the scan time 26 Chapter 4 Basic Data Acquisition Benchtop System Users Manual Release 4 3 When the Stepping Mode box is checked the stage moves incrementally an image is acquired while the stage 15 stationary the stage moves incrementally again another image is acquired while the stage is stationary and so on This behavior is preferrable when the scanning speed is very slow when the integration time is very long or to guarantee there is no motion blur in your scan Scanning Speed is the linear speed of the stage during a scan If the Go Home After box is checked the stage will return to its starting position after a scan The speed at which the stage returns to its original position is the Homing Speed and the Jog Speed is used for the Jog Stage buttons described above 4 4 Focus Objective Lens You are now ready to focus the objective lens of your Pika imaging spectrometer At first this process is somewhat challenging but with a little practice it becomes quite easy Begin by clicking on the Focus button located on the Sp
51. ant Senescence Reflectance Index Red Edge Normalized Difference Vegetation Index Output Files Simple Ratio Index p M 1 a Select output file location omlbatch processor testlresults Transformed Chlorophyll Absorption Reflectance Index Place output files in same folder as input files Vogelmann Red Edge Index 1 Vogelmann Red Edge Index 2 Copy LCF and times files Export render KMLs E Add completed cubes to workbench Use the directory tree to select the cube plugins you wish to run Cube plugins generate new datacubes and will run in the order selected as shown in the Cube plugin sequence panel Each plugin operates on the output of the plugin before it Each plugin s setup window appears as you add it to the processing queue In the figure above the Airborne Radiance Conversion plugin is set to run followed by the NDVI plugin Select any render plugins you wish to run Render plugins generate images from the final result of the cube plugin sequence and appear in the Renders panel You can choose to save renders as png tiff or jpg with Render Export Format the figure above no render plugins are selected so no images will be generated Use the controls at the lower right to select the location to save results Copy LCF and times files and Export render KMLs are useful for our airborne customers If desired the resultant datacubes can be added
52. arams asFloat scale the datacube from 0 1 bool asBIP force the data to be in BIP interleave lines samples bands Additionally the function util makeEmpt yCube is convenient for constructing new datacubes It s usage is as follows from resonon core data import util newcube util makeEmptyCube mode memory typechar f rotationString datacube getRotationString 10 7 Example Plugins The following plugin examples are based on built in plugins from Spectronon Warning Each plugin in Spectronon needs to have a distinct class name If you change the name of the following plugins don t just remove Example from the class name or your plugin will conflict with a built in plugin 10 7 1 Cube Plugins Cube plugins are the most common They operate on a datacube and return a datacube Cube plugins add themselves to the New Cube submenu and context menu A very simple example based on a Spectronon built in plugin is shown below from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecWavelength Benchtop System Users Manual Release 4 3 from resonon core data import util class BandRatioExample CubePlugin Create a new single band cube that results from divide two bands of an existing cube The plugin s help text label Band Ratio Example name of plugin as it appears in Spectronon default rendering others include SingleBand ThreshToColor Tr
53. as an image file Clicking on this option will open a window that allows you to name the file select the image format and save the plot image file in a location of your choosing This option is available for the Spectral Plotter X Cross Section plot and the Y Cross Section plot 7 6 3 Save as Text This tool allows you to save the plot data values as a txt file Clinking on this option will open a window where you can name the file for the txt file and save it in allocation of your choosing This option is available for the Spectral Plotter X Cross Section plot and the Y Cross Section plot 7 6 Plots Menu 75 Benchtop System Users Manual Release 4 3 7 6 4 Clear This tool allows you to remove a spectral plot To use this tool select the Plot you wish to remove in the Resource Tree then select Plots Spectral Plotter Clear 7 7 Datacubes and Header Files Datacubes collected with Resonon imaging spectrometers are ENVI compatible ENVI is a standard software pack age for hyperspectral data analysis that can be obtained from ITT http www ittvis com ENVI compatible data requires two files one that contains the hyperspectral data and also a header file that contains important information about the data Hyperspectral data can be accessed with Matlab documentation for opening multi band image files in Matlab can be found here http www mathworks com help techdoc ref multibandread html 7 7 1 Hyperspectral data files
54. ative vectors in the middle of each cluster to perform SAM classification it is easy to see there would be substantial misclassification E g two red points lie nearly along the blue representative vector 82 Chapter 8 Advanced Data Analysis 2 Hyperspectral Classification Benchtop System Users Manual Release 4 3 Red Blue For the case shown above a different algorithm is needed One approach that would clearly be superior to SAM at least for these data is to find the center points of each set of training points and then all points within a certain radius would be classified as that object This is also a well known hyperspectral classification algorithm known as Euclidean distance Again this approach is readily extended to hyperspectral data because the concept of distance between points is well established for multi dimensional space Red Blue The Euclidean distance approach can readily be improved upon by recognizing that a circle is not the best enclo sure and an ellipse whose axes were scaled to the point distribution width would be far better The hyperspectral classification algorithm that utilizes this information on the distribution of training points is called the Mahalanobis distance approach As one would expect the cost of additional algorithm sophistication is often the need for additional processing There are of course many other approaches to classifying objects using hyperspectral data such as
55. ce This tool allows you to convert your datacube to either radiance or reflectance values as described below 66 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 Spectronon Pro File Image Spectrum Selection Plots Spectrometer Window Help New Image aln end E 500 Info New Cube Utilities r Res New Stack Classify Oc A s nalyze gt 1 Save Cube lyz Correct r Correct From Cube Save Cube As Color Correct From Spectrum AS Rename Cube XC E Agricultural gt Radiance Conversion Close Cube Custom User written Plugin Reflectance Conversion Datacube New Cube Correct Correct From Cube This tool allows the user to scale their datacube to another known reference datacube When you select this option a Correct From Cube tab will appear in the tool control panel In the Correction window select the cube you wish to correct to Typically this is a datacube you have recorded of a uniform reference material Once you click Update a new datacube will be generated that is scaled to the reference datacube If you click Float Mode data brightness will be scaled to 1 and thus if your reference has a reflectance of 1 your data will be in units of reflectance Otherwise the data will be scaled to the maximum bit level of your reference cube E g a pixel whose brightness matches a 12 bit reference datacube will be set to 212 4096
56. d in Chapter 5 The Stop Focus Tool option turns off the live view which you need to do before you record a scan The last live view will remain in a tab of the image panel Another way to turn off the live view is to click the Focus button Fd which toggles live view off and on The Record Dark Current Cube is used to remove dark current noise from the imager When you select this option you will be instructed to block the imager and then click OK Another way to do this is select the Dark Current button D on the SpectrononPro toolbar 53 Additional Controls 0 43 Benchtop System Users Manual Release 4 3 Spectronon Pro T File Datacube Image Spectrum Selection Plots Window Help E y TAR d Reload Imager Info X Reload Stage Resource Tree e Cubes Auto Exposure A Spectra Start Focus Tool Ae Calibration Stop Focus Tool Record Dark Current Cube Remove Dark Current Cube Set Dark Current Cube Record Response Correction Cube Remove Response Correction Cube m Set Respose Correction Cube No Tree Selection Start Recording Stop Recording Fig 5 11 Spectrometer menu The Remove Dark Current Cube option will remove your dark current calibration cube This will also remove the red check from the Dark Current button gl The Set Dark Current Cube allows you to import a saved dark current correction cube This option is rarely utilized The Record Response Correction Cube option allows you to use a reference mate
57. demagnify tool and the pan tool in the spectral plotter to navigate and examine features 6 7 Saving Spectra Plots and Images Spectronon makes it easy to save the results of your work for further investigations or for making presentations To save a spectrum click the spectrum you wish to save in the Resource Tree and then you may either 1 use the Spectrum menu in the main menu or 2 right click on the spectrum in the Resource Tree to reveal the menu shown below From this menu select either Save Spectrum or Save Spectrum As This will open a save dialog Once saved the new name will appear when the file is plotted in the spectral plotter and the file can be re opened for use in later sessions 58 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 Spectronon Pro File Datscube Image e mie L4 Sire Spectre Save Spectrum amp s Ooh Spectrum Hide Plot Sek Label Color Copy Spectrum Ac Text Irfo SPECTRUM LISTED IN RESOURCE TREE Select the menu option Copy Spectrum As Text to copy the data onto your clipboard from which you can paste it into other applications such as Notepad and Excel Spectronon Pro Fie Datacube Image Spectrum Selection ele spectrometer Window Help d e 4m a a E il Spectral Plotter Fl Save as Image Info X Cross Section FM _Save as Text Cross Section Clear o Resource Tree To save a plot us
58. does 7 2 Datacube Menu 67 Benchtop System Users Manual Release 4 3 NOT account for spatial variability such as vignetting and slit variations the data should be converted to radiance before using this plugin The radiance conversion plugin is described below To utilize the Correct From Spectrally Flat Reference tool it is best if there is a reflectance reference target located within your hyperspectral image First re a L boned use one of the ROI tools marquee or lasso or and generate a mean spectrum Then select Datacube New Cube Correct Correct From Spectrally Flat Reference Enter in the percentage reflectivty O 1 of the reference material If you click Return Ints 0 1000 the values will be scaled to 1000 Otherwise the data will be scaled to 0 1 and returned in floating points Datacube New Cube Correct Radiance Conversion This plugin requires a radiometric calibration datacube and will return radiance units of microflicks To use apply the plugin to the cube to convert and select the appropriate radiometic calibration datacube If camera dark current is significant an optional dark cube can be collected at the appropriate integration time and gain and utilized Select this cube if necessary Unlike some other plugins the Return Floating Point checkbox will return the same values microflicks as the integer mode but may be necessary for very bright scenes if radiance values above 65
59. e the sample number int line the line number int asFloat scale the datacube from 0 1 bool LEC ELAS ID array of length bands getBandNumForWavelength self wavelength return the band number nearest the given wavelength params wavelength wavelength of light in nanometers float returns 2d array getBandCount self return The number of bands in this cube aka the number of different wavelengths getSampleCount self return the number of samples for the cube aka the image height getLineCount self return the number of lines for the cube aka the image width 10 6 Using Datacubes 109 Benchtop System Users Manual Release 4 3 def def getBandAtWavelength self wavelength asFloat False return band at given wavelength params wavelength the wavelength of the band you want float asFloat scale the datacube from 0 1 bool returns the band for the given wavelength frequency 2d array return the band with the closest wavelength to the given value in the case of equidistance between two bands the lower of the two bands is returned getBandWithName self name asFloat False return band with given name params name the name of the band you want float asFloat scale the datacube from 0 1 bool FOLUEns the band with the given name 2 array you may pass the string form of a known wavelength
60. e 4 3 ARROWS SHOW THE 4 CANDIES ila Eu Gg E pos TT m Eko Ha a FROM WHICH ROIS WERE ifo X pe TED l i Te SELECTED TO GENERATE THE 4 A e ose SPECTRAL CURVES SHOWNM IN THE OP orange Yellow Pas O Orange PC SPECTRAL PLOTTER Spectra cec spex BH A Calbr ation Corina Info s Cube Ci Documents and Update Seething Ban Mis or n B ge veles FACE bun Contrast Presets A Pier Test Pil Clase Triana an BGB 3 irae From 3 gren bands of a cube True Color Filter Tool Contract Class Stretch Coler infrared Phe bigi wahr between Pen Qeon per cente 40 0 amounts Bea Ji 69556 5 FHS Data ions 45 0 isan 693 53 04 4525 4 x 10 14 030 OFS The four spectral curves shown in the spectral plotter will be used as the Reference Spectra to perform a SAM clas sification To perform the classification click on Datacube New Cube Classify Spectral Angle Mapper SAM Spectronon Pro File PESENE Image Spectrum Selection Plots Spectrometer Window Help New Image rR oo Cube k ilitie ites Classify Analyze Correct New Stack Re c7 Save Cube f Save Cube As Close Cube Color Agriculture Euclidian Distance Quadratic Discriminant Analysis Spectral Unimix L k S
61. e 4 3 Tabs will appear in the Parameters window that allow you to change the image more on this later Header information on your datacube will appear in the information panel Note Spectronon can open any datacube with an ENVIO formatted header This includes bip bil and bsq formats This chapter employs an example datacube of M amp M and Reese s Pieces candies This datacube can be down loaded from Resonon s website at http downloads resonon com By default the datacube is opened with a true color image of the data which approximates the appearance of the object under normal lighting conditions by combining red green and blue wavelengths from the datacube Spectranon Pro men Orange Yellow Plri RGH Tok X Resource Tree Spectrogram Y cubes Orange Yellow Pt bip 1 ML ossi OPENED FILE LISTED HERE is ecra Calibration 0 6 4 02 x 02 04 06 1 Contras Info Cubs CADocusente and e Auto Update Settings Aandi eskan ME Or an l a Presis P Render Tao AaB less TriBand enun R33 3 angpen mage ror rr P li IE AT IMAGE FROM Fitar Tool Contrast Class Stretch DATACUBE FILE Infrared Stretch the brightness values z beberen percentage 4 amp 40 0 amatz Red J 404 695 693 555 994 4625 Data a 550 0 Green J 404 5385 693 8505 904 4625 i H 460 0 TOOL CONTROL PANEL HHGS cud SS FHS
62. e Sample Bin and Line Bin allow you to bin pixels along a spatial x y axis The Sample axis refers to the cross track axis of the imaging spectrometer and the Line axis refers to the along track axis of the imaging spectrometer The Spectral Bin slider allows you to bin spectral channels and is likely the most useful of the binning options After clicking Update a new binned datacube is generated 7 2 Datacube Menu gg Benchtop System Users Manual Release 4 3 Note The binned data are averaged and thus do not change significantly in amplitude If you choose Float Mode the data will be rescaled with 1 set to the maximum bit level of the data If your data have been scaled to a reflectance reference you will find that glints can create Brightness values larger than one Datacube New Cube Utilities Crop Wavelengths Generates an image in the Image panel and a new tab with sliders in the tool control panel that allow you to crop wavelength bands by choosing a new minimum and maximum wavelength within the cube Once chosen click the Update button Hint You may want to click on the RGB tab in the tool control panel to reset the bands used to create the image after cropping wavelengths Datacube New Cube Utilities Subtract Spectrum Generates a new cube by subtracting a background spectrum from all pixels in the datacube This option is useful for example if you are monitoring fluorescent dyes and wish to subtract the
63. e steps may have already been completed for you First enable jumbo packets for your network adapter in Windows In Windows click the start menu right click on My Computer and clicke the Manage option Click on Device Manager in the left column click on Network Adapters in the middle column right click your wired network adapter and choose Properties In the Advanced tab click Jumbo Packet and set the Value to 9014 Bytes Click OK and close the Computer Management window In Spectronon two additional settings will appear under the subgroup GigE Properties in the Pika II tab under the File Preferences menu Packet size and Packet Delay These settings allow for management of bandwidth between the camera and the controlling computer The settings selected here will in part determine the achievable framerate Larger packet sizes require less CPU usage to transfer a given amount of data from the camera to the computer than small packet sizes The maximum packet size is limited by your network adapter any other network hardware and the camera Typically the largest available packet size 1s recommended 36 Chapter 5 Advanced Data Acquisition 5 1 Camera Controls Benchtop System Users Manual Release 4 3 Integration Time Use Extended Integration Range On Camera Binning Spectral Binning Adjust window on change Spectral Calibration slope 0 54 intercept 378 87 Fig 5 4 Pika XC imager tab in the Pref
64. e the Plots menu as shown below Select which plot you wish to save Spectral X Cross Section or Y Cross Section and then select Save as Image to save as an image or Save as Text to save the plotted data as tables in text file Both options will pop up a save Dialog To Save an Image select mage from the main menu and then Export Image This will pop up a save dialog Spectronon Pro File Datacube a Export Image PES Close Image Spectrum Selec 6 8 More Information Complete information about Spectronon Tools and hyperspectral classification can be found in the Benchtop System Users Guide located at the product support page at the Resonon website http www resonon com customer_support html 6 8 More Information 59 Benchtop System Users Manual Release 4 3 60 Chapter 6 Basic Data Analysis SEVEN ADVANCED DATA ANALYSIS 1 GENERAL A complete listing and explanation of Spectronon s data visualization tools is presented in this chapter The organiza tion of this chapter follows the main menu buttons 7 1 File Menu 7 1 1 Open Datacube Clicking this option will open a window that allows you to browse and select the datacube you wish to work with Only ENVI compatible files will appear Once selected click Open and the datacube will be loaded and an image will be generated in the image panel Spectronon Pro Datacube Image Spectr Open Datacube
65. ectance Index 2 CRI2 NIR Red Enhanced Vegative Index ENPe RM Modified Chlorophyll Absorption Reflectance Index MCARI Pero 0 249 700 P cso s70 Modified Red Edge Normalized Vegetation Index MRENDVI Pro Pm _ Prso Pros 2 Daas Paas Pon NIR Red NIR Red Pos P sro Photochemical Reflectance Index 22 21 Pos Pro ae Peso Psoo Plant Senescence Reflectance Index PSRI oa Poso Pros Poso Pros Modified Red Edge Simple Ratio Index MRESR Normalized Di Ter ence Vegelalive Index NDVI Red Edge Normalized Difference Vegetation Index RENDVI NIR Simple Ratio Index SR Red Pass Structure itive Pi t Ind Ld Se Pana Peso Transtormed Chlorophyll Absorption Reflectance Index TCARI 2 Peso psa 2 B70 Vogelmann Red Edge Index 1 Vogelmann Red Edge Index 2 Vogelmann Red Edge Index 3 Povo Water Band Index WEI PER S0 9 1 Introduction 91 Benchtop System Users Manual Release 4 3 9 2 Generating HVI maps in Spectronon The datacube in the following example was acquired using a Resonon airborne hyperspectral imaging system This datacube is available for free download from http downloads resonon com in Spectronon can be found at the path Datacube New Cube Agriculture
66. ectrononPro tool bar This will reveal a live image from the camera within your Pika imaging spectrometer Wave your hand in the field of view of your Pika imaging spectrometer to confirm that the image is a live view One axis of this image represents the spatial position axis of your object and the other is the spectral wavelength axis To understand this view better move colored objects within the field of view of your imager after you have focused the objective lens Place an object with multiple light and dark regions within your Pika imaging spectrometer s field of view A sheet of paper with dark lines as provided in the chapter titled Focusing amp Calibration Sheets works well If you are in the field and are observing objects at a distance direct your Pika towards an object with multiple features such as a tree with many branches Unless your lens is already focused you will see a series of blurry or barely discernable lines in the Image Panel of SpectrononPro TYPICAL BLURRY LIVE VIEW IMAGE SPECTRAL AXIS SPATIAL AXIS To adjust the focus first unlock the focus adjustment With Schneider lenses this is done by loosening the locking metal collar on your objective lens using an Allen wrench size 5 64 inch 4 4 Focus Objective Lens 27 Benchtop System Users Manual Release 4 3 LOOSEN LOCKING COLLAR TO ENABLE FOCUSING Then rotate the objective lens until you see dark lines from your object come into focus as
67. election options are located in the workbench tab of the preferences window A selection transparency of 1 represents a 52 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 completely transparent selection the selection will not be visible while a selection transpatrency of 0 represents a completely opaque selection the underlying render of the datacube will not be visible through the selection Warnings Warn on Unsaved Last Scan Scan Settings Detauit San Image 6 5 Image Visualization Hyperspectral data can be visualized in far more ways than conventional color images Image controls are provided in the tool control panel By default the image is displayed in True Color which means three representative bands are used to generate a red green blue RGB image approximating how it appears to a human eye The Color Infrared preset option provides a false color RGB image with the red band set to an infrared wavelength This option is useful for live vegetation datacubes Any time you wish to restore the image to true color simply click on the True Color button under Presets 6 5 Image Visualization 53 Benchtop System Users Manual Release 4 3 Contras gt Auto Update _ RGB Contrast Presets True Colar Color Infrared gt 640 0 404 6305 699 5805 934 4625 i ss Green a 404 6385 600 5805 994 4575 SLIDERS ALLOW YOU TO CHANGE THE BANDS USED TO
68. elf my_options value choice two pass action for choice two elif self my options value choice three pass action for choice three 10 4 5 SpecCube class SpecCube Spec Displays a ComboBox for selecting one of the loaded datacubes def init self label datacube wb requireMatchedLineCount False requireMatchedSampleCount False requireMatchedBandCount False requireBandCount None defaultValue None The SpecCube user interface is a combo box that will automatically be populated with the datacubes that are currently available on the Spectronon workbench e g any cube that appears in the Spectronon resource tree The parameters passed to SpecCube s constructor allow you to provide criteria by which the list of applicable cubes will be filtered 104 Chapter 10 Custom Data Analysis Writing Plugins Benchtop System Users Manual Release 4 3 e if requireMatchedLineCount True the list will only show cubes whose line count is equal to that of the cube passed as datacube e ifrequireMatchedSampleCount True the list will only show cubes whose sample count is equal to that of the cube passed as dat acube e if requireMatchedBandCount True the list will only show cubes whose band count is equal to that of the cube passed as datacube e if requireBandCount an integer the list will only show cubes whose band count is equal the spec ified integer This value overrides that provided in requireMatched
69. enchtop System Users Manual Release 4 3 Spectronon Pro File Datacube Image Spectrum Selection Plots Spectrometer Window Help Bh cus 0 ine Orange Yellow PWOS RGBE Took Resource Tree Spectrogram X jd ED cubes orange Yellow PW09 bip 4000 Orange Yellow PWO9 RGE INTENSITY PLOT FOR RGB Spectra 3200 X Calibration 2400 1600 5 Conkrals Info F Auto Update Cube C Documents and Settings Rand Desktop Meds Oran RGB ge Yellow Pwo bip Contr ast Presets 5 Render Tool RGB Class Triband return an RGB 3 layer image from True Color 3 given bands of cube Filter Tool Contrast Class Stretch Color Infrared Stretch the brightness values between two give percentage 4 640 0 amounts 404 6985 699 5805 994 4625 Data samples 438 T site ine 343 INSPECTOR TOOL LOCATION Green J ar 404 6985 699 5805 994 4625 23 amp 460 0 RGB VALUES OF PIXEL AT INSPECTOR TOOL LOCATION T 404 6985 699 5005 994 4625 Clicking on the X and Y tabs in the plot panel accesses the corresponding cross sectional plots The plot will show the intensity versus position value for the RGB bands used to create the image or the Grayscale band if used with a grayscale image Note The direction of X and Y depends on the orientation of your cube Moving the Inspector Tool should reveal which axis you are plotting Hint Use the magnify tool the
70. ength in the spectral plotter and also lists the curve in the Resource Tree 74 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 7 5 7 Mean Second Derivative This tool plots the second derivative of the mean spectral curve within the ROI as a function of wavelength in the spectral plotter and also lists the curve in the Resource Tree Hint You may want to close all other spectra before using Mean First and Second Derivative tools as the magnitude of the derivative values will typically be very different than the values of the spectra themselves 7 5 8 Send to Clipboard This tool allows you to transfer the data within your selected ROI The first option is Copy Mean as Text which allows you to paste the mean spectrum values into other applications such as Notepad or Excel The second option Spectra as Text copies the spectral curves of all the pixels within your selected ROI This tool is useful for those who want to perform specific statistical analysis on the data from small regions 7 6 Plots Menu The Plots menu items provide tools for the data shown in the spectral plotter Spectrometer Window Help Spectral Plotter 4 Set Range X Cross Section 4 Save as Image Y Cross Section 4 Save as Text Clear Copy to Clipboard As Text 7 6 1 Set Range This allows the user to set the range of the X axis and the Y axis 7 6 2 Save as Image This tool allows you to save the plot
71. entific computing in Python using Numpy arrays Many Spectronon features make use of SciPy libraries and SciPy is available to the user for use in developing additional plugins To develop a Spectronon plugin create a file with a py extension such as MyCustomPlugin py and place it in the folder SpectrononAppData user_plugins This folder is typically located at C Users lt your user name gt SpectrononAppData user_plugins In the file write a subclass of the plugin type you desire as described below You can use any text editor of your choice to create the plugin a list of Python text editors can be found here Upon loading Spectronon will scan the user plugins floder for py files and if a valid plugin is found add it to the Spectronon menu system 10 2 Spectronon Plugin Overview There are 4 types of plugins CubePlugin RenderPlugin 99 Benchtop System Users Manual Release 4 3 SelectPlugin FilterPlugin built in plugins only Plugins are creating by defining a subclass of one of the base plugin types then placing the source file that defines that plugin in the spectrononAppData user plugins folder The base plugin type definitions must be imported from the spectronon workbench plugin package and subclassed Each plugin class has three methods that you can override to define your custom behavior For example from spectronon workbench plugin import CubePlugin class MyPlugin CubePlugin The docstring
72. erences window 37 Benchtop System Users Manual Release 4 3 The following properties are available for this network adapter Click the property you want to change on the left and then select its value on the right Value Adaptive Inter Frame Spacing 901 4 Bytes Enable PME Energy Efficient Ethemet Flow Control Interupt Moderation Intemupt Moderation Rate 4 Checksum Offload Large Send Offload IPv6 Link Speed amp Duplex Link Speed Battery Saver Locally Administered Address Fig 5 5 Device Manager Network Adaptor Jumbo Packets The packet delay sets the gap between data packets transmitted to the computer from the camera and allows for time for the CPU to process the data it has received before new data is transmitted If your scans result in dropped frames try increasing the packet delay which will impact the maximum framerate achievable Typically problems with these settings will manifest as image inconsistency errors which are reported in the bottom status bar as a scan progresses See the figure below for an example warning Inconsistency warning Try increasing packet delay 5 2 Stage Controls You can move the stage manually by clicking on the Jog Stage buttons located on the tool bar of SpectrononPro The buttons will move the stage incrementally in either direction Use the buttons to center the stage underneath the Pika imaging spectrometer 38 Chapter 5 Advanced Data Acquisit
73. ethings e andi Oran Mode Wavelengh ge Yellow POS bip io 5 Bases Render Tool RGB Class Triband retum RGB 3 Isver mage from Red J 3 oven Bands of cube 404 6285 899 5605 aos 462 Fiker Tack Contrast Clas 4 F 516 72366 Stretch the brightness values between bun percentage Green amens WH GRS ed 4625 4 455 2135 Data ee ae Bhia samples J G e ine 825 404 528 99 5005 gos 4627 305 Presets Ge 1906 B 214 True Color Color Inifrsred Note The Red Green and Blue vertical lines in the Spectral Plot show the location of the bands chosen to create the false color RGB image 6 5 Image Visualization 55 Benchtop System Users Manual Release 4 3 Controls Auto Update RGB Contrast Use Contrast Enhancement Presets 2 stretch Linear stretch Adjustments lt 2 Low Cut Percentage 50 100 4 gt 98 High Cut Percentage al o0 100 Inverse 0 all Individual The Contrast tab in the tool control panel allows you to adjust the image contrast If Use Contrast Enhancement is not checked no image enhancement will be done and the tools in the Contrast tab will be not be active Note Contrast enhancement does NOT change the hyperspectral data It only changes the way the image appears Generally contrast enhancement is beneficial The 2 stretch is the default and it sets the darkest 2 of the pixels in the ima
74. first select a representative set of training 79 Benchtop System Users Manual Release 4 3 pixels from the image of each of the objects of interest This might be done for example by selecting the pixels within each grey rectangle indicated in the image above A useful way to visualize the color information in these training pixels is to plot the red and blue brightness values for each selected pixel in color space along blue and red axes as shown below Note This color plot uses only the training pixels selected within the three grey rectangles shown in the image above Red Blue We can immediately recognize the training pixels that align along the vertical red axis are associated with the red ellipse as those pixels clearly have large red brightness values but small blue brightness Similarly the blue triangle pixels lie primarily along the horizontal blue axis The purple pixels however lie in between the two axes because the color purple has significant red and blue brightness 1 it is a mixture or red and blue Because some of the pixels are dark and others are light the distribution of training pixels from each object is spread out from small values to large values in a near linear manner One can see that the pixels from the spectrally distinct objects are separated in this color space Although this example has a 2 dimensional color space one could create a 3 dimensional color space for RGB cameras or 10
75. g spectrometers are line scan instruments Thus by adjusting the step size of the scanning system you are adjusting the spacing of the lines used to assemble your image To adjust your image s aspect ratio it is useful to image an object whose distortion is easy to observe such as a circle For example print out Pixel Aspect Ratio Calibration Sheet provided in the Focusing amp Calibration Sheets chapter of this manual Place an object with circles within the field of view of your Pika imaging spectrometer and record a scan with enough lines that you can see the complete circle You may need to record several trial images to determine how many lines to scan If your image is distorted along the scan direction change the scanning speed on the Stage tab on the Tools panel If your image is elongated along the scan direction decrease the scanning speed conversely if your image is too short along the scan direction increase the scanning speed After resetting the steps per scan record a new image to observe the change in image distortion Repeat the above process until your image is no longer distorted 4 6 Scanning and Saving Datacubes To record a hyperspectral datacube image put the number of lines you wish to scan in the window to the left of the Scan button gt Then press the Scan button to record a datacube A waterfall image of your datacube will appear in the Image Panel of Spectronon and a new entry labeled Current Scan
76. ge to a value of 0 and the brightest 2 of the pixels in the image to maximum brightness 255 This choice minimizes the impact of glare You can customize the percentage of the dark pixels set to 0 and the percentage of the bright pixels set to 255 with the sliders The Linear stretch option sets these percentages to zero The nverse checkbox is useful if you wish to highlight dark pixels The ndividual checkbox controls whether the brightness levels of the three image layers are considered all together or as individual layers Spectronon Pro File EET eats Image Spectrum Selection Plots cs NES New Stack Ser aner Re e Save Cube ff Save Cube s Close Cube 56 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 It is often useful to view a single band in a standard grayscale black and white image to visualize the impact of a single spectral feature To do this go to the main menu and select Datacube New Image Grayscale The controls are similar to the RGB controls If Auto Update is not checked be sure to click Update after moving the slider to see the grayscale image for a new band Spectronon Pro File Ostacube Image Spectrum Selection Picts Spectrometer Window Help Jb Oe OC 0 Inf Tods x Cranes PIMIO R GE Drange Yellow PW S Greyscale Resource Tres Specbrogeam x Y ae cube El orange Yellow POS bip
77. he Target Pixel parameter is at least as bright as the Target Brightness but is not saturated For example if the Target Pixel is equal to 100 and the Target Brightness is equal to 3890 the default settings for 12 bit cameras the algorithm will attempt to find an exposure time such that the brightest pixel has a brightness of at least 3890 but less than 4095 saturation Similarly if the Target Pixel 95 the pixel that represents the 95th percentile of total brightness will be targeted in each frame instead of the brightest pixel Selecting a Target Pixel less than 100 may cause the brightest pixels to be saturated but may be useful for achieving desired exposure when features of interest are not the brightest in the field of view For typical conditions Target Pixel is set to 100 and Target Brightness is set to 95 of the camera s bit depth 5 1 4 Windowing and Binning of Hyperspectral Data Windowing refers to using a reduced subset of the available data which reduces overall data volumes Windowing can occur in both the spectral and spatial dimensions Both the Pika XC and Pika NIR have windowing capability while the Pika II does not For the Pika NIR windowing can also be used to eliminate portions of the focal plane array not utilized by the imager which are the wavelengths outside the 900 1700 nm range 5 1 Camera Controls 35 Benchtop System Users Manual Release 4 3 Binning involves combining data from multiple adjacent p
78. he objective lens blocked click OK as instructed SpectrononPro will then collect multiple dark frames and use these measurements to subtract the dark current noise from your measurements The Dark Current button on the toolbar will appear with a red check through it as soon as the dark frames have been collected T Once you see the red check unblock the objective lens 4 5 2 Set Reflectance Reference Measuring absolute reflectance of an object requires calibration to account for illumination effects To do this click on the Response Correction Cube button R on the SpectrononPro toolbar A message will appear telling you to place a reference material within your Pika imaging spectrometer s field of view The reference material should be uniform across the imager s field of view Examples of reference materials include Spectralon or sheets of white Teflon Please place a white target in front of the camera and click Ok or Cancel to skip Response Recording Once the reference material is in place click on OK This will trigger a short scan of the reference material Once complete the Response Correction Cube will appear with a red check mark e indicating that the data you collect will be scaled in reflectance to your reference material including flat fielding to compensate for spatial variations in your lighting Note The Dark Current button gi and the Response Correction Cube button ee are disabled in live camera view mode
79. he plugin This method returns the results of your plugin Cube plugins must return a datacube from their action method Render plugins return an image representation of a datacube as a numpy array 100 Chapter 10 Custom Data Analysis Writing Plugins Benchtop System Users Manual Release 4 3 e Select plugins do not return a value but interact directly with the workbench Filter plugins cannot be written by users If you want to get information from the user you also define a setup method and define some Specs more on specs below within it def setup self define some specs here If your UI can be altered based on user entry into other UI elements for example the user selects how many inputs they want to provide then you must write an update method def update self add or remove specs or update paramaters of existing specs None of these methods accepts arguments Instead there are attributes of the Plugin that you can use 10 3 The attributes available to each plugin type self wb Plugins This is a handle on the spectronon workbench The workbench gives you a lot of freedom so use this reference wisely Some common uses of self wb are e popup dialog messages such as self wb postMessage The combination of arguments is invalid for this cube retrieve data from the workbench correctioncube self wb getCube self correctCubeID value plot something to the plotter self wb plot m
80. headerfile exists save header file to disk to reflect new larger dimensions of cube appends a band and increases number of bands in the cube by one getSubCube self minsample 0 minline 0 maxsample 1 maxline 1 mode memory interleave None get SubCube return a cube with the data in the given range mode and interleave with no arguments this will return a complete copy getFramelessCopy self makeTypeFloat False mode None asBIP False return an empty cube object with a header based on the header of this cube optional makeTypeFloat will return getBandlessCopy self makeTypeFloat False mode None return an empty cube object with a header based on the header of this cube optional makeTypeFloat will return newcube self getEmptyCubeWithHeaderCopy makeTypeFloat mode newcube setMetaValue bands 0 return newcube setBand self bandnumber band writes given band at given band Benchtop System Users Manual Release 4 3 params bandnumber the number of the band to overwrite returns None setFrame self line frame write a given frame into location at given line params line the line number to overwrite int frame the content of the frame 2 array returns None This is only for replacing frames that are already there to add Frames to a cube use appendFrame def getArray self asFloat False asBIP False get the numpy data array for this datacube p
81. iBand etc defaultRenderer SingleBand allows control for what which users the plugin is available 1 is for normal users userLevel 1 def setup self GUI controls that allows the user to select wavelengths as they exist in datacube self numband SpecWavelength label Numerator Wavelength datacube self datacube self denomband SpecWavelength label Denominator Wavelength datacube self datacube def action self make an empty cube for the results preserving metadata from the old cube newcube util makeEmptyCube mode memory typechar f rotationString self datacube getRotationString get the data array from the datacube at the two specific wavelengths topband self datacube getBandAtWavelength self numband value asFloat True bottomband self datacube getBandAtWavelength self denomband value asFloat True perform the math on the numpy arrays result topband bottomband put the results in the empty cube newcube appendBandWithName result bandname self label return the new cube and it will be added to the Spectronon file tree return newcube 10 7 2 Render Plugins Image Plugins Render Plugins create a 2D representation of a datacube known as an Image in Spectronon They operate on a cube and return a 2D image of or fewer bands or colors A simple example is shown below import numpy from spectronon workbench plugin import RenderPlugin class B
82. ika XC or Pika NIR The options for different Pika models will vary 33 Benchtop System Users Manual Release 4 3 Open Datacube Open Spectrum Preferences Built in Scripts Run Script Exit 0 00215 10 00556 20 00898 0 0 eee 0 0 12 01768 2403535 Spectral Calibration slope 10 nm per band 00 Fig 5 2 Pika II imager tab in Preferences window 34 Chapter 5 Advanced Data Acquisition Benchtop System Users Manual Release 4 3 5 1 1 Basic Camera Controls Integration Time also known as Exposure Time is the duration of data acquisition for each individual line image If you use the Auto Expose button in the toolbar the integration time will be adjusted automatically for your lighting conditions such that there are no saturated pixels in the image Make certain the reflectance calibration tile is being imaged when using this option Frame Rate is equal to the number of images acquired each second and limits the maximum integration time Max Integration Time seconds 1 0 Frame Rate If you do not have enough light decreasing the frame rate will enable a longer integration time Gain is a factor which increases the signal but at the expense of Signal To Noise ratio in units of dB We recommend that you operate at low Gain values preferrably zero to minimize noise However for situations where you need to operate at high frame rates and need more sign
83. instead of 2 Thus the size of your datacube will double with Float as compared to not using this option When checked the Update Plots option allows the plots to update during focusing and scanning This option should normally be checked unless you have a sluggish data acquisition computer and you require more speed 5 3 3 Spectrometer Menu The Spectrometer Main Menu button provides alternative approaches to useful actions as well as some timesaving options The Reload Imager option will restart the Pika imager without restarting SpectrononPro software This option will save you time if you have for example accidently disconnected your Pika spectrometer or crashed the software The Reload Stage option will restart your scanning stage without restarting SpectrononPro software This option saves you time if your stage needs to be restarted The Auto Exposure option allows you to let the software set the exposure setting for your lighting conditions To use this place an object within the field of view of your Pika and then click on Auto Exposure Alternatively you can E use the Exposure button E from the toolbar The Start Focus Tool option will provide you with a live view from your Pika imager in the image panel This tool is useful for adjusting the focus of your objective lens or checking the illumination of your system You can also start the focus tool by clicking on the Focus button LF on the SpectrononPro tools bar as describe
84. ion Benchtop System Users Manual Release 4 3 ay Plotters Scanner workbench Pika Stage slope 1 0244 intercept 394 77 Camera Settings 10 529245 24 00007 GigE Properties Packet Size 576 4 k 0 Fig 5 6 Pika imager tab of the Preferences window WARNING 1502 frame inconsistencies JOG STAGE BUTTONS File Datacube Image Spectrum Selection Plots Spectrometer Window Hel ba OS Oe ae 06 0 6 gt P fps Fig 5 7 Jog Stage buttons located on the tool bar 5 2 Stage Controls 39 Benchtop System Users Manual Release 4 3 Further control of the stage can be controlled from two windows the Stage tab in the Tools Panel and the Preferences window accessed by File Preferences Tools Controls Camera Stage Linear Speed Units Rotation Motor Stepping Mode Scanning Speed cm s 0 07938 7 97719 15 875 Go Home After Homing Speed 0 cm s 0 07938 7 97719 15 875 Jog Speed p cm s 0 07938 7 97719 15 875 Fig 5 8 Stage tab in Tools Panel Speed Units is a setting used for different types of stages Linear is used for the standard linear translation stage that is installed on most benchtop systems Rotation is used for a tripod mounted rotational scanning stage typically used in outdoor applications Motor displays speed in motor pulses per second Stepping Mode controls the way the stage moves in relat
85. ion to the imager and is ideal for very slow scanning very long integration times or to guarantee there is no motion blur in your scan Scanning Speed is the linear speed of the stage during a scan If the Go Home After box is checked the stage will return to its starting position after a scan The speed at which the stage returns to its original position is the Homing Speed and the Jog Speed is used for the Jog Stage buttons described above 40 Chapter 5 Advanced Data Acquisition Benchtop System Users Manual Release 4 3 5 3 Additional Controls 5 3 1 Scanner Preference Tab The Scanner tab allows you to set the binning options for recording your datacubes These options are useful for situations where you wish to reduce the size of collected datacubes or where you wish to improve your signal to noise ratio The Spectral Bin option indicates how many spectral channels to bin For many situations this is your best choice to reduce the size of your datacubes and to improve your signal to noise ratio as the spectral resolution is often higher than needed With the Pika II a spectral bin of 3 will result in 80 spectral bands 240 3 The Spatial Bin option indicates how many spatial channels to bin This option is typically useful when imaging an object known to be largely homogeneous Q Preferences amp 2 Scanner Workbench Plotters Pika NIR Stage Binning Spectral Bin 4 pixels Spatial Bin tjt pixels
86. is similar to SpecCube but allows for choice of a spectrum object from the workbench e if requireMatchedWavelengths True the list will only show spectra for which the wavelengths of the spectrum exactly match those of the cube passed as dat acube e if requireMatchedBandCount True the list will only show spectra for which the number of bands matches the number of bands of the cube passed as dat acube Most commonly you will pass your plugin s datacube member to SpecCube s constructor as the datacube argument but any datacube object is a valid choice You should always pass the plugin s self wb workbench member to SpecCub as the wb argument The value of a SpecSpectrum is an index to a spectrum object within the workbench To get the spectrum object itself you must call wb tree get Spect rum method Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecSpectrum class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my_spectrum SpecSpectrum label Select a Spectrum datacube self datacube wb self wb requireMatchedWavelengths True requireMatchedBandCount False defaultValue None self my_spectrum help my mouse hover help def action self spectrum self wb tree getSpectrum self my spectrum value operate on the spectrum object 10 4 7 SpecWavelength class SpecWavelength SpecFloat def
87. ixels into a single data point Binning reduces overall data volumes and increases SNR at the expense of decreased resolution Binning is possible in both the spatial and spectral dimensions There are two types of binning One type is software binning which occurs in Spectronon and is available on all systems The other type is hardware binning which is only available on the Pika XC In addition to decreasing data volumes and increasing SNR hardware binning when applied to the spectral dimension only has the additional advantage of increasing the maximum frame rate See the table below for details Pika XC Spectral Binning Parameters Spectral Bin 4 of Spectral Spectral Max Frame Factor Channels Resolution nm Rate Fig 5 3 Hardware Binning Parameters for the Pika XC On the Pika XC it is preferrable to bin with hardware than with software which will utilize fewer CPU resources Software binning is available on the Scanner tab in the Preferences window see section 7 3 1 below Windowing and hardware binning controls are accessed through the appropriate imager tab in the Preferences window The Start Spectral and End Spectral values determine which spectral bands are collected The Start Spatial and End Spatial values determine which spatial bands are collected GigE Connection Setup The Pika II imager with GigE connection requires some additional steps for complete setup If you purchased your computer from Resonon thes
88. k Update This will update your image Use the inspector tool e to see the AE values in the Data panel To obtain a visual representation of the AE values click on the Threshold to Colormap tab in the parameters window and adjust the thresholds Agriculture This option allows the user to create maps of Hyperspectral Vegetation Indices HVI s HVI s in Spectronon are discussed in greater depth in Chapter 9 Clustering Clustering tools are unsupervised classification routines that do not require any a priori information of a spectral image They group the scene into clusters based on spectral similarity Supervised classification routines can almost always produce better results if scene knowledge available 7 2 3 New Stack This tool enables you to overlay multiple images This tool is particularly useful for presenting classification results For example we may have a True Color RGB image of an object such as shown below as well as a classification map of one of the candy types Using the Stack tool we can combine these images to show the classification map on top of the RGB True Color image 7 2 Datacube Menu gg Benchtop System Users Manual Release 4 3 af TRUE COLOR IMAGE CLASSIFICATION MAP THE STACK TOOL COMBINES MULTIPLE IMAGES CLASSIFICATION MAP STACKED ON TRUE COLOR To use the Stack tool follow the path Datacube New Stack This will open a Stack tab in the tool control panel Use the slider
89. k microflicks are expected Datacube New Cube Correct Reflectivity Conversion This plugin is intended for data collected with the Resonon airborne hyperspectral imaging system It requires a spectral imager radiometric calibration cube downwelling irradiance data relevent to the datacube being collected and downwelling irradiance sensor radiometric calibration Optionally dark current data for the spectral imager and downwelling irradiance sensor can be used if the dark current for these systems is significant The returned data will be in units of reflectivity from 0 1 for float mode or 0 1000 in integer mode Anomaly Anomaly detection tries to determine anomalous outliers in a datacube that do not conform to the expected spectra This is usually based on a datacube of clutter which contains the expected spectra Color The Color option allows you to transform your hyperspectral data into CIE colorspace providing XYZ xyY and LAB values for each pixel Additionally you can determine the AE values for each pixel as compared to a standard set by the user Spectronon Pro File Image Spectrum Selection Plots Spectrometer Window Help TAN New Image aa E 500 Info New Cube gt Utilities gt Res New Stack Classify c gt Save Cube Mme c Correct Save Cube As Color gt CIE Colorspace Conversion Z i b Jel C Close Cube Agricultural Delta E Custom User
90. kbox change the interface type set the interfaceType member of the spec Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecBool from resonon constants import INTERFACES class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my bool SpecBool 1 1 Option default True self my bool units my units self my bool help my mouse hover help self my bool interfaceType INTERFACES CHECKBOX action self 1f self my_bool value 102 Chapter 10 Custom Data Analysis Writing Plugins Benchtop System Users Manual Release 4 3 pass action if true else pass Faction if false 10 4 2 SpecFloat class SpecFloat SpecNumber E Creates a slider or spinner for setting a floating point number def init self label minval maxval stepsize 1 defaultValue None The default interface type of a SpecFloat is a slider A spin button can also be used Example from spectronon workbench plugin import CubePlugin from resonon utils spec import SpecFloat from resonon constants import INTERFACES class MyPlugin CubePlugin An example plugin label My Plugin userLevel 1 def setup self self my float SpecFloat label My Float minval 0 maxval 100 stepsize 0 1 defaultValue 10 4 self my float units my units self my float help my mouse hover help
91. l Approach 81 Benchtop System Users Manual Release 4 3 Red Blue By finding the orange reference vector and establishing the threshold of acceptable angles from this vector we have trained our imaging system To identify objects with the same color as the purple parallelogram one images the objects and then calculates the angle as described above for each pixel Pixels with an angle smaller than the threshold are identified to be the same material color as the purple parallelogram One could also train the system to find pixels the same color as the red ellipse or blue triangle by finding the appropriate reference vector for these objects thereby enabling one to identify multiple objects within one image by calculating the angle between multiple object reference vectors To extend this approach to hyperspectral data with multiple spectral channels one utilizes the more general definition of a vector dot product where the sum is over all spectral bands dimensions indexed by i The angle 0 x y is still the angle between the two vectors and P x y although now in a complex multi dimensional color space that is more difficult to visualize the mathematics and approach are identical Of course SAM is not always the best algorithm to use as one can see with the example below where pixels from two hypothetical objects we wish to distinguish between are shown as red and blue points In this case if one choose represent
92. l the software it will be necessary to set the appropriate version flags To do so create an executable shortcut on the desktop and right click on the shortcut to bring up the Properties window On the Shortcut tab in the Target field add the following flag relevant for your imager to the executable name 3 2 Setup of Hyperspectral Imager 21 Benchtop System Users Manual Release 4 3 22 Chapter 3 System Assembly FOUR BASIC DATA ACQUISITION 4 1 Start The System If you have a lighting system turn on the lights and let them warm up It may require 15 to 20 minutes for the illumination to fully stabilize pectrononPro With the camera and scanning system connected to your computer launch SpectrononPro software by double clicking on the SpectrononPro icon or starting SpectrononPro from your Start menu The SpectrononPro user interface is shown below with the various windows labeled 23 Benchtop System Users Manual Release 4 3 RESOURCE TREE PANEL MAIN MENU TOOLBAR IMAGE PANEL PLOT PANEL Imag Plots Spectrometer Window 12 amp E File Datacube cti Help leaf smali RGB TENE Resource Tree m Spectra X Y e Cubes leaf small bip 6000 3 leaf small RGB A Spectra 4800 Calibration 3600 2400 1200 Resource Info Cube C Program Files SpectronomPro 920 400 480 560 640 720 800 880 960 lexamples Veaf_small bip Render Tool
93. le Rename Spectrum Close Spectrum Show Plot HidePlot Set Label Color Show Region Copy Spectrum As Text Crop Spectrum 7 4 2 Save Spectrum As This tool allows you to save a spectrum in a filename of your choice 7 4 3 Export Spectrum as Text File This tool allows the user to save a spectrum in a text file 7 4 4 Rename Spectrum This tool allows you to rename a spectrum 72 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 7 4 5 Close Spectrum This tool will close the spectrum selected in the Resource Tree and remove the plot from the spectral plotter 7 4 6 Show Plot This tool will plot the spectral curve in the spectral plotter selected in the Resource Tree Typically this option is used after you have hidden the plot with the following Hide Plot option 7 4 7 Hide Plot This tool will hide the spectral plot shown in the spectral plotter To use this tool select the spectrum you wish to hide in the Resource Tree and then click Hide Plot 7 4 8 Set Label Color This tool allows you to change the color of spectral curves as well as classification maps generated with the spectral curves To change colors of spectral curves select the spectral curve you wish to change in the Resource Tree Clicking on Set Label Color will reveal a pallet of colors Choose a new color and then click OK and the color will be reset 7 4 9 Show Region Use this tool to reveal the
94. lected tolerance Increasing the tolerance will result in a larger selected region with greater spectral variability within that region i e it allows pixels that are less similar to the clicked pixel to be included in the selection Decreasing the tolerance will result in smaller selected regions with greater spectral similarity As with the other selection tools holding control while using the floodfill tool will allow a selection to be built up through multiple clicks of the tool 50 Chapter 6 Basic Data Analysis Benchtop System Users Manual Release 4 3 An ROI consisting of multple parts of the image has been selected by using the floodfill tool several times One of the most useful selection options is mean spectrum Descriptions for the other ROI options can be found in Chapter 4 Selecting the mean spectrum option calculates the mean spectrum of all the pixels within the ROI area you selected and plots the result in the spectral plotter Plots Spectra X Y 15000 Assorted plastics spec 0 Assorted plastics spec 2 7 13200 11400 LF FIqnine Li Wo e e e 73800 600 900 960 1020 1080 1140 1200 1260 1320 1380 1440 1500 1560 1620 1680 mm Wavelength inm 6 4 Region Of Interest ROI Tools 51 Benchtop System Users Manual Release 4 3 Individual spectra can be selected by clicking on the graph of the spectra You can crop an individual spectrum by selecting it se
95. lecting a rectangular region in the plot window right clicking to bring up the menu and selecting crop spectrum You can also set the range of the plot by selecting Plots Spectral Plotter Set Range Hint To examine the plots in more detail you may resize the plot panel boundary by dragging the edges Alterna tively click on the magnify tool then click or drag in the spectral plotter to zoom in The pan tool will allow you to pan within the spectral plotter as well Selecting Mean Spectrum creates a new entry in the resource tree under a new heading spectra Right clicking on the spectrum in the resource tree will reveal a menu of options Some of the most used options are listed below Spectronon Pro Sa File Datacube Image Spectrum Selection tr Save Spectrum Save Spectrum As Export Spectrum As Text File Rename Spectrum Close Spectrum Show Plot Hide Plot Set Label Color Show Region Copy Spectrum As Text Crop Spectrum Save Spectrum Change the name and save the spectrum as a file Set Label Color Open a color picker dialog to change the label color of the spectral plot and as shown later classification areas based on this spectrum Show Region Show the originally selected area for the ROI in the current image This option is useful after you have selected several ROIs The color and transparency of the selection can be modified by selecing preferences from the file menu The s
96. mager and lighting assembly are adjustable along the length of the tower See Figure 1 below Resonon s hyperspectral imagers are line scan imagers also referred to as push broom imagers Two dimensional images are constructed by translating the sample relative to the camera This is typically accomplished by placing the sample on a linear translation stage HYPERSPECTRAL IMAGING CAMERA MOUNTING TOWER SPECTRONONPRO SOFTWARE ON DESKTOP COMPUTER POWER SUPPLY FOR LIGHTING ASSEMBLY LIGHTING ASSEMBLY LINEAR TRANSLATION STAGE Fig 2 1 Benchtop hyperspectral imaging system Resonon s Pika imaging spectrometers are compact high fidelity digital instruments for industrial and scientific ap Benchtop System Users Manual Release 4 3 plications Spectronon is a powerful hyperspectral data visualization and analysis software package we provide as a free download Spectronon is easy to learn offers efficient workflow and is highly extensible by the user for custom applications Additionally a number of datacubes can be downloaded from the Resonon website www resonon com so you can begin exploring hyperspectral data within a few minutes SpectrononPro has all the features of Spectronon but also includes data collection tools that are highly integrated with our Pika imaging spectrometers to streamline the collection of spectral images 8 0 0 0 0 0 00000 Chapter2 System Overview
97. n provides visualization and manipulation capabilities for hyperspectral images SpectrononPro software has all the features of Spectronon but also enables data acquisition from Resonon s family of imagers Spectronon software can be downloaded for free on Resonon s website http www resonon com SpectrononPro comes bundled with any of Resonon s imaging spectrometers For more advanced analysis capabilities software packages such as ENVI are available This chapter begins with basic operation of Spectronon such as opening a hyperspectral datacube and viewing the data A complete description of visualization tools is provided in Chapter 8 Advanced Data Analysis Chapter 9 Custom Data Analysis Writing Plugins discusses how to implement user written algorithms into Spectronon enabling custom data analyses References to Spectronon apply to SpectrononPro as well 6 1 Spectronon Tools To open a datacube select File Open Datacube Spectronon Pro Datacube Image Spectr Open Datacube Open Spectrum Preferences Built in Scripts 4 Run Script Exit This will open a dialog that allows you to browse to find your datacube Select your datacube and click on Open to load your datacube This will result in the following e An image of your datacube will appear in the image panel A listing of the open datacube will appear in the Resource Tree 45 Benchtop System Users Manual Releas
98. oduced or transmitted in any form or by any means electronic or mechanical including photocopying for any purpose without written permission from Resonon Resonon will not be responsible or liable for any accidental or inevitable damage that may result from unauthorized access or modifications This document may contain errors or inaccuracies and it may be revised without advance notice This manual is updated frequently Please see our software copyright and EULA at http www spectronon com docs Spectronon_EULA pdf Resonon welcomes any recommendations regarding this manual Customer feedback is always welcome as it helps us to continuously improve upon the quality of our products Please contact us at Resonon with any questions or comments We look forward to hearing from you 127
99. of the HVI at any pixel can be obtained with the Inspector Tool eo File Datacube Image Spectrum Selection Plots Window Help A ele fo In DCL s VNIR airborne datacube 1 RGB VNIR Resource Tree 6 Cubes g VNIR airborne dataci GRAB THE INSPECTOR TOOL HERE 9 2 Generating HVI maps in Spectronon 93 Benchtop System Users Manual Release 4 3 The value of the HVI along with the coordinates of the inspector tool can be found in the lower left corner of the screen Spectronon File Datacube Image Spectrum Selection Plots Window Help b enses into VNIR airborne datacube 1 RGB VNIR airborne datacube 1 ARI2 Greyscale Resource Tree OF Cubes VNIR airborne datacube 1 1 airborne datacube 1 RGB VNIR airborne datacube 1 ARI2 1 VNIR airborne datacube 1 ARI A Spectra X Calibration Resource Info Cube VNIR airborne datacube 1 ARI2 Render Tool Greyscale Class SingleBand Return a single band rendering from one of three selection techniques Filter Tool Contrast Class Stretch Stretch the brightness values between two give percentage amounts sample 726 line 1277 Anthocyanin Reflectance Index 2 0 112816 Another tool for visualizing HVI maps is to transform the grey scale image into a colormap This can be accomplished with Datacube New Image Scalar to Colormap 94 Chapter 9 Advanced Data Analysis
100. ollowing simplified hypothetical examples In general the algorithms used to classify objects scale with the number of spectral channels Understanding how the general approaches work with a simple 2 color system makes it relatively easy to understand hyperspectral classifica tion with tens or hundreds of channels Therefore consider a 2 color camera that provides digital numbers for how much red and how much blue is in each pixel Thus this example is even easier than a conventional digital color camera that provides digital values for the three colors red green and blue Much like humans need to learn to distinguish between objects classification algorithms generally also must be trained Typically this is done by imaging samples of interest and then using results from this training set to learn how to distinguish between objects in general As an example assume we wish to distinguish between the three objects shown below a red ellipse a blue triangle and a purple mixture of red and blue parallelogram Ignore the grey rectangles for now An x and y axis is drawn to provide location coordinates for each pixel Note that each object has a distribution of light and dark pixels although each object is approximately the same color If we image these objects with our hypothetical 2 color imager that senses only and blue channels there will be two spectral channels per pixel To train our imaging system we
101. pability to support user defined data analysis algorithms commonly called plugins Writing plugins allows the user to extend Spectronon s functionality by incorporating custom tools designed for their specific applications Furthermore plugins can create interactive user interfaces and integrate directly into Spectronon s menu system Warning Plugins have direct access to much of Spectronon s internal code We make this functionality available in the hope that it will be a useful tool Use of this tool is at the user s own risk Resonon cannot support every situation that may arise from the use of this information The methods and objects documented here may change at any time Only the methods and attributes documented here are intended to be used in custom plugin development all other methods and properties of the workbench plugin datacube and spec classes are for internal use only and should not be called or overridden 10 1 Getting started Spectronon plugins are written in the Python programming language and a working knowledge of Python is a pre requisite for writing Spectronon plugins If you re new to programming or to Python the Beginner s Guide to Python provides a collection of useful resources for getting started Additionally familiarity with the computing tools pack age Numpy is important as Resonon datacubes are based on Numpy arrays Finally the SciPy library provides an excellent collection of tools for sci
102. pectral Angle Maper SAM Li k Spectra X m This will open a new window as shown We wish to classify 4 objects so use the slider or arrow keys to select a Spectrum for each box 8 2 Spectral Angle Mapper SAM Classification 85 Benchtop System Users Manual Release 4 3 Controls Update Auto Update SA Contrast Layers J Spectrum a Spectrum Spectrum Spectrum This will bring up 4 Spectrum pull down menus Click into each one of these and select one of the 4 spectra created with the ROI tool Then click OK and Spectronon calculates the spectral angle as described above for each pixel for all 4 Reference spectra Typically several seconds are required for the calculation which generates a new classification map in the image panel using a default set of Threshold values 86 Chapter 8 Advanced Data Analysis 2 Hyperspectral Classification Benchtop System Users Manual Release 4 3 Controls Update Auto Update Contrast Threshald To Colormap Mone m3 0 0912 Thresh 0 J 0 0213 0 7198 1 4183 ay 0 0925 Thresh 4 J 0 0229 0 7185 1 414 0 0902 Thresh 2 J 0 0203 0 7196 1 419 ED 0 0905 Thresh 3 J 0 0207 0 7181 1 4155 Adjust the Threshold values to obtain more accurate classification by clicking on Threshold Colormap tab in the tool control panel To adjust the Thresholds move the sliders select the a
103. peso and are shown graphically in the figure below 89 Benchtop System Users Manual Release 4 3 Mean Spectrum 600 700 wavelength nm HVI s have found many applications in research using airborne remote sensing data and are currently subjects of intense research efforts 9 1 1 Useful References The developers at ENVI have written an excellent description of HVI s which is available at their website We at Resonon consider ENVI owned by Exelis to be excellent hyperspectral data analysis software Here is a link to their HVI website https www exelisvis com Learn WhitepapersDetail TabId 802 ArtMID 2627 ArticleID 13742 Vegetation Analysis Using Vegetation Indices in ENV I aspx Another very good reference discussing HVI s is this book Thenkabail Lyon and Hute Remote Sensing of Vegetation CRC Press 2011 9 1 2 HVI s in Spectronon calculated by Spectronon are listed in the following table Also listed are the equations that Spectronon uses for each HVI 90 Chapter 9 Advanced Data Analysis 3 Hyperspectral Vegetation Indices Benchtop System Users Manual Release 4 3 Hyperspectral Vegetation Index Equation in Spectronon 1 Anthocyanin Reflectance Index 1 pm 1 Anthocyanin Reflectance Index 2 NIR Red y Blue Red NIR Red y Blue Red Atmospherically Resistant Vegatative Index Carotenoid Reflectance Index 1 Carotenoid Refl
104. relatively large for the red and blue pixels Thus if we calculate the angle between the reference orange vector and all pixels in the image we recognize that those pixels with a small angle are from the purple parallelogram and those pixels with a large angle from the orange reference vector are not purple pixels Fortunately there is an easy way to calculate the angle between two vectors by utilizing the vector dot product To do this write the orange reference vector in component form as R Ry where the subscript r indicates the red brightness value component and b indicates the blue brightness value compo nent Similarly for all pixels in the image write P x y 2 P x y where y indicate location of the pixel in the original image with an ellipse triangle and parallelogram The vector dot product of and any pixel P x y is Re RP x y R P a y cos 0 a y g R R R RyRp 2 and x y is the angle where Rland P x y are the gt magnitudes of H and P x y E between the reference vector R and the pixel vector P x y ReP Pn R P x y By choosing only those pixels that have an angle 0 x y less than some small threshold value one can effectively identify all the purple pixels Graphically this is equivalent to choosing only those pixels within a narrow green cone as shown below Solving for 0 2 y yields 8 1 Genera
105. rial against which all measurements will be scaled Typically this is done with a reference material whose reflectivity is approximately equal to 1 When you select this option you will be instructed to place a reference material within the imager s field of view and then click on OK A small reference cube will then be recorded and used for scaling the data to the bit level of the camera E g for the Pika II 12 bits or 4096 and for the Pika NIR 14 bits or 16 384 If you have selected Normalize to Float as described above the data with the scaled to 1 rather than the bit level of the camera Another way to perform this function is to use the Response Correction Cube button on the SpectrononPro toolbar The Remove Response Correction Cube will remove the existing response correction cube This will also remove the red check from the Response Correction Cube button e The Set Response Correction Cube allows you to import a saved reference cube This option is rarely utilized as one should generally record a correction cube regularly and with the current lighting to obtain accurate results The Start Recording and Stop Recording options will start and stop a scan with your Pika imaging spectrometer This can also be done using the Scan Button gt on the SpectrononPro toolbar The Scan Button will toggle Start and Stop functions 4A 0 Chapter 5 Advanced Data Acquisition SIX BASIC DATA ANALYSIS Spectrono
106. rrow keys or enter values by hand Each Reference spectrum has its own threshold After adjusting a threshold click Update and a new classification map will be generated in the Image panel With a few tries a classification rendering similar to the one shown below can be generated Only those pixels within the threshold are colored the classification colors match those shown in the spectral plotter and Resource Tree for the Reference spectrum If a pixel s Spectral Angle is within the threshold for more than one Reference spectrum the quantity Spectral Angle Threshold Value is calculated for each Threshold and classified for the class that minimizes this quantity This weighting scheme allows you to emphasize or deemphasize each class to fine tune your classification map 8 2 Spectral Angle Mapper SAM Classification 87 Benchtop System Users Manual Release 4 3 Spectronon Pro Fle Datacube Image Spectrum Selection Pio Spectrometer Window Help Dette 0 Info Took x Grange Yeh Pahl Drange Yellow Threshold To Colormap Resource Tres Spectrogram X Y gj cubes er Grange Tebow Pais Dip a 4000 _ _ Orange Yek PWISRG CLASSIFICATION MAP nam r3 n mcg Erro Pul Fao E _ id E Orange fellow POS 2400 on j 1800 spec 1 per 9 gna ails 900 S00 GUI fU BED SDDO IPHI 2 lav amp 4 Controls info Au
107. s from the bottom of the post Benchtop System Users Manual Release 4 3 REMOVE BOLTS FROM THE BOTTOM OF THE POST Align the post to the baseplate then tip the baseplate up and attach using the four bolts just removed from the bottom of the post d bu E ro E CB rw i A nc A After tightening all four bolts tip the tower upright Connect the lighting assembly to the regulated power supply 12 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 REMOVE THE HOOK AND LOOP TABS HOLDING THE WIRING TO THE LIGHTING ASSEMBLY REG LATED DC POWER s ses e25 VWSAMLEY FNIT Cer PUSHING THE WIRE IMTO TOWER SLOT PROTECTS IT FROM THE LAMP HEAT Plug the regulated power supply into a 120 V socket and turn on to test the lighting Turn the lights off after testing to continue with assembly 3 1 Assembly of Stage Tower and Lighting 13 Benchtop System Users Manual Release 4 3 PLUG IN REGULATED POWER SUPPLY AND TURN ON RED POWER SWITCH TO TEST THE LIGHTING If the lights do not turn on turn off the red power switch and check all connections If the lights continue to fail contact Resonon Note When illuminating your sample with Resonon s lighting system allow 20 minutes for the lighting to fully stabilize Using the supplied USB cable connect the stage to your computer Plug the cable into a black USB 2 0 port not a
108. s the command quantities in brackets are acceptable arguments for the commands and statements in square brackets are additional explanation Note that the order of these commands can be different from file to file ENVI All ENVI compatible header files begin with this interleave bil bip bsq Tells the software the hyperspectral data format data type 4 12 4 is 4 byte floating point 12 is 2 byte unsigned other data types are also supported but these are common lines The number of lines in the hyperspectral data where is a number samples The number of samples in the hyperspectral data where is a number This is typically the number of cross track spatial pixels of the hyperspectral imager bands The number of bands in the hyperspectral data where is a number 76 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 bit depth 12 14 The bit depth of the hyperspectral data shutter The shutter time in units of msec where is a number This is mostly for reference but it may be used by certain plugins gain The camera gain in dB where a number This is mostly for reference but it may be used by certain plugins framerate The framerate at which the data were recorded in frames sec where 15 a number This 15 mostly for reference but it may be used by certain plugins reflectance scale factor
109. tacube you will see a history header value that shows the way this cube was generated pointlist long list of values This is the set of points from the original cube that were averaged together to make this spectrum if the spectrum is a mean label color FF00FF This is the color used to plot the spectrum 7 7 Datacubes and Header Files 77 Benchtop System Users Manual Release 4 3 78 Chapter 7 Advanced Data Analysis 1 General EIGHT ADVANCED DATA ANALYSIS 2 HYPERSPECTRAL CLASSIFICATION 8 1 General Approach The detailed spectral information in hyperspectral data enables one to distinguish between very similar objects Ad ditionally machine vision algorithms enable rapid accurate and repeatable classification of objects With these capabilities hyperspectral imaging has a broad range of current and potential applications including sorting food raw materials recycled materials quality control pharmaceuticals food printed goods remote sensing defense search amp rescue mineral exploration agriculture to name just a few Critical to all of these applications is implementation of algorithms that classify the pixels within an image based on their spectral profiles This chapter provides an introduction to hyperspectral data classification There are many approaches to classifying objects However the general approach to most of these algorithms can be understood by considering the f
110. text will appear on the user interface as help text when hovering over the plugin s menu item label My Plugin userLevel 1 def setup self wee The setup method is where you will create the plugin s user interface You can also define any custom initialization behavior here Do not override CubePiugin Init Ji def update self The update method is called each time the user changes the value of a Spec You might need this for your plugin s book keeping def action self The action method is called when the plugin is run and should contain the logic of your custom behavior Always define a label attribute as above which will become the text that is displayed in Spectronon menus The docstring used for the plugin class will become the help text when the user hovers over your custom plugin s menu item Always define userLevel 1 at the top of the class definition Occasionally a user may use user level 2 which will render the plugin invisible unless Spectronon is invoked in expert user mode Warning Failure to set the userLevel to 1 or 2 at the top of the plugin definition will prevent the plugin from attaching to Spectronon s menu system Writing any plugin requires at minimum that you write an action method def action self your stuff here The action method is called by Spectronon when the plugin is activated and again each time the user interacts with any Specs defined by t
111. the one indicated 8 1 General Approach 83 Benchtop System Users Manual Release 4 3 below where one draws a line between the two sets of training points The system is trained by noting that all pixels that map to the left of the line are associated with one object whereas the pixels that map to the right of the line are associated with the other Extrapolating this approach to higher dimensions is more difficult as the line becomes a plane in three dimensions and a hyper plane in color spaces with dimensions larger than three Red Blue 8 2 Spectral Angle Mapper SAM Classification In this example different kinds of M amp M and Reese s Pieces candies are classified using the same hyperspectral datacube shown in Chapter 5 This datacube can be downloaded from Resonon s website To perform SAM refer ence spectra must be collected for the objects of interest In this case based on prior knowledge we know that the four candies indicated below see arrows in image are all different Using the marquee tool or lasso select small ROIs on each of the four candies avoid the glare spots indicated in the figure below After selecting the ROI right click and then select Mean Spectrum This will generate four spectral curves in the spectral plotter and also list the four spectral curves in the Resource Tree as shown below 84 Chapter 8 Advanced Data Analysis 2 Hyperspectral Classification Benchtop System Users Manual Releas
112. the portion of the array that has been selected the pointlist is given in order sample line subarray cube_array self pointlist 1 self pointlist 0 calculate the average of the selected region ave subarray mean display the calculated result for the user self wb postMessage ROI Average s ave title Average ELEVEN BATCH PROCESSING Manually running the same processing sequence on a series of datacubes can be a tedious process Spectronon s batch processor allows the user to choose a number of datacubes at once and set up a sequence of processing steps plugins to execute on the chosen files This may be of particular interest to users of our Airborne product but applies to any customer who needs process multiple datacubes at once To start the batch processor select File Batch Processor The batch processing window appears Batch Processor Select the datacubes to be processed H E SWTOOLS i sor_test test 1 1 20150527_19h37m_pika_gige_1 bil EE Users sor_test test data Area 1 2120150527 19h38m gige 2 bil UR iis sor testitest data Area 1 3120150527 19h38m pika gige 3 bil P tom gimp 2 8 matplotlib PyCharm4 VirtualBox ef results test data ff Area 1 1 Hf Area 1 2 EE Area 1 3 fl bin fl Contacts E E Desktop fl Documents fl Downloads H E Favorites E E Links aE Music ff Pictures ff Saved Games
113. to Up Cube Crane Til Pe L Auto Updabs Fir Maler am JAH Threchele To Doksani Render Tool Threshold To Cokin Cles Trine Tocoler rebum an mae From 4o 0 0912 3 green barsss of a cube Thresh Ji Tool Contrat Cle Sirei Ne 1 463 View the athusl values vt out T F Cuan Theesht Dei DO AUS Le a Nona X ob Thresh 02185 DT ps 419 oos THRESHOLDS USED TO GENERATE os Oana 0 7101 1 4155 SAM is only one of many many possible classification algorithms Other classification algorithms are accessed in a similar manner as described in Chapter 8 section 2 2 2 These algorithms can be found by clicking Datacube New Cube Classify Additionally user defined scripts can be written and used with Spectronon for custom classifications algorithms 88 Chapter 8 Advanced Data Analysis 2 Hyperspectral Classification ADVANCED DATA ANALYSIS 3 HYPERSPECTRAL VEGETATION INDICES 9 1 Introduction Hyperspectral Vegetation Indices HVI s are a common tool for analyzing hyperspectral data of agricultural crops They are typically simple arithmetic combinations of sums and differences of spectral reflectance values at particular wavelengths For example the Structure Insensitive Pigment Index SIPI is defined by the following expression SIPI P800 0445 Ps00 0680 where the values p445
114. to select how many images you wish to combine Stack height Then click Update Pull down menus appear that allow you to select the images you wish to combine slider Alpha is available for each image that allows you to set its transparency in the combined image Once set click Update and your combined image will appear Tools for saving and closing datacubes are under the Datacube main menu option 7 2 4 Save Cube This tool will save the datacube currently open Since many datacubes can be open at the same time be sure to select the cube you wish to save in the Resource Tree before clicking on Save Cube If the datacube has not been previously saved a window will open that allows you to rename the datacube and save it in a file of your choosing 70 Chapter 7 Advanced Data Analysis 1 General Benchtop System Users Manual Release 4 3 Spectronon Pro File MEE Image New Image k New Cube b Info Mew Stack Save Cube Save Cube s _ Close Cube 7 2 5 Save Cube As This tool allows you to save the currently open datacube under a new name or to a new location Clicking on this option opens a window that allows you to name the datacube and save it in a file of your choosing 7 2 6 Rename Cube Use this to rename datacubes 7 2 7 Close Cube The tool closes the datacube selected in the Resource Tree If this datacube has not been saved a message will appear asking to confirm whether or not
115. to the workbench to show in Specrtonon when processing is finished Press the OK button to begin processing A progress dialog appears Benchtop System Users Manual Release 4 3 Batch Processor Batch Processing Complete 16 46 59 Opening cube C Users tom batch_processor_test test data Area 1 1 20150527 19h37m_pika_ 16 47 00 Running plugin Airborne Radiance Conversion 16 48 50 Running plugin Normalized Difference Vegetative Index NDVI 16 48 50 Saving result to C Users tom batch processor test pesults 20150527 19h37m_pika_ gige 16 48 50 Opening cube C Users tom batch_processor_test test data Area 1 2 20150527 19h38m_pika_ 16 48 51 Running plugin Airborne Radiance Conversion 16 50 53 Running plugin Normalized Difference Vegetative Index NDVI 16 50 53 Saving result to C Users tom batch processor testiresults 20150527 19h38m gige 16 50 53 Opening cube C Users tom batch_processor_test test data Area 1 3420150527 19h38m 16 50 53 Running plugin Airborne Radiance Conversion 16 51 34 Running plugin Normalized Difference Vegetative Index NDVT 16 51 34 Saving result to C XsersYtomlbatch processor testlresults 20150527 19 gige 16 51 34 Batch Job Complete Press the Close button to dismiss the dialog after processing is complete The results are saved to the selected location 117 Benchtop System Users Manual Release 4 3 TWELVE FOCUSING amp CALIBRATION SHEETS
116. trum Preferences Built in Scripts Run Script Exit Locate the value for Slope on the calibration sheet provided with your Pika imaging spectrometer Enter this value in the window labeled slope in the Preferences menu and then click on set Locate the value for Intercept on the calibration sheet provided with your Pika imaging spectrometer Enter this value in the window labeled intercept in the Preferences menu and then click on set Preferences oB as Scanner Workbench Pika NIR Plotters ENTER SLOPE AND CLICK SET Spectral Calibration ENTER INTERCEPT AND CLICK SET slope 5 447 nm per band intercept 644 74 nm Initial setup is now complete Instructions for how to collect hyperspectral images are given in the chapters on data acquisition 3 2 3 Objective Lens Installation Resonon imaging spectrometers are designed to operate with Schneider c mount lenses To install the Schneider lens first remove the lens and imager from their packaging Unlock the set screw on the chrome lens collar of the Schneider lens with a 2 mm hex wrench NOTE Do NOT remove the screw all the way just loosen it a few turns 18 Chapter 3 System Assembly Benchtop System Users Manual Release 4 3 X TA A Holding onto the objective lens by chrome collar unscrew the black lens barrel out until there is approximately a 10mm gap between the barrel and collar M
117. will appear in the Resource Tree To save the scanned datacube image use your mouse to select Current Scan and then either right click or select Datacube Save Cube This will open a new window that allows you to name the datacube and save it in a folder of your choosing 30 Chapter 4 Basic Data Acquisition Benchtop System Users Manual Release 4 3 Spectronon Pro File Datacube Image Spectrum Selectii 162 AMA Info lt Resource Tree cy Cubes New Image 4 A Spectra New Cube X Calibration New Stack Save Cube Save Cube As Rename Cube Close Cube If you do not save your datacube the Current Scan will be overwritten when you record another datacube A warning will appear 4 6 Scanning and Saving Datacubes 31 Benchtop System Users Manual Release 4 3 32 Chapter 4 Basic Data Acquisition FIVE ADVANCED DATA ACQUISITION 5 1 Camera Controls Control of the imager can be controlled from two windows the Camera tab in the Tools Panel and the Preferences window accessed by File Preferences Tools Controls Camera 5tage Camera Settings daas 53 17 Framerate p hz 0 04 31 07 62 11 ha ul 7 809 J e 0 01 50 005 100 0 0 Gain dB 0 17 34 Fig 5 1 Camera tools in Preferences window When the Preferences window opens select the tab for your specific Pika imager Pika II P
118. yarray and there are specs SpecCube SpecSpectrum that expect the workbench as a parameter to help them work their magic For more information see Using the Workbench self datacube CubePlugin RenderPlugin SelectPlugin This is the datacube on which this plugin will be applied Cube plugins action methods are expected to return a new datacube based somehow on this one Render plugins will use this datacube as a data source to build a viewable image For more information see Using Datacubes self pointlist SelectPlugin For any selection this is a list of points that was inside the selected area in the form of a numpy array of shape number of points 2 where self pointlist 1 sample line index of the ith selected point 10 4 Specs A Spec is a class that manages a piece of data and a user interface to that data They underly much of Spectronon s codebase A Spec maintains the connection between the user interface and the data in the back end When a plugin class has a spec member Spectronon will generate an appropriate graphical widget to allow the user to adjust that spec s value in the plugin s control panel Specs can be imported from resonon utils spec Usually you will access the spec by getting the current value in an action method using the spec s value member value mySpec value Here is the list of available Specs 10 3 The attributes available to each plugin type 101 Benchtop System Users Manual Release

Benchtop System Users Manual

Contents

Download Pdf Manuals

Related Search

Related Contents