Home

5. Feature Tracking & Visualization in VisIt

1. 91 Pick Modes VisIt provides a way to interactively pick values from the visualized data using the visualization window s Zone Pick and Node Pick modes When a visualization window is in one of those pick modes each mouse click in the visualization window causes VisIt to find the location and values of selected variables at the pick point When Visit is in Zone pick mode it finds the variable values for the zones that you click on When Vislt is in node pick mode similar information is returned but instead of returning information about the zone that you clicked on VisIt returns information about the node closest to the point that you clicked Pick is an essential tool for performing data analysis because it can extract exact information from the database about a point in the visualization You can put the visualization window into one of VisIt s pick modes by selecting Zone Pick or Node Pick from the Popup menu s Mode submenu After the visualization window is in pick mode each mouse click causes VisIt to determine the values of selected variables for the zone that contains the picked point or the node closest to the picked point Each picked point is marked with an alphabetic label which starts at A cycles through the alphabets and repeats The pick marker is added to the visualization window to indicate where pick points have been added in the past To clear pick points from the visualization window select the Pick point option from
2. lt PluginName gt MDServerPluginInfo C h lt I SI S S A lt PluginName gt ScriptingPluginInfo C h v lt PluginName gt ViewerPluginInfo LC h v Table 2 List of auto generated files in VisIt 46 1 Use xml2plugin tool under src bin directory to generate the plugin framework The commands for using xml2plugin are as follows cd lt visithomedirectory src operator FeatureTrack gt bin xmi2plugin FeatureTrack xml Incase VisIt s bin folder is added to your system path then just type xml2plugin FeatureTrack xml 47 2 Copy the files as mentioned in Table 2 in this folder To make your operator function properly you need to modify some of the auto generated files Since we are using most of the work for this operator from previous section 23 instead of modifying auto generated files we just use files from previous work 3 Now compile and FeatureTrack operator plugin is ready to be used The following commands are used to compile a operator cd lt visithomedirectory src operator FeatureTrack gt make Caution Whenever a plugin has to be installed which is written by someone else and not included in VisIt The above steps should be followed Merely copying all the files into the new folder by the name of the plugin and compiling that plugin won t work because the environment is different on machines Machines can have different version of Qt VisIt and
3. en wikipedia org wiki Map C 2B 2B 25 Observer Design Pattern http userpages umbc edu tarr dp lectures Observer pdf 26 VisIt installation https wci llnl gov codes visit source html 27 S Y Chen and X Shan High Resolution turbulent simulations using the Connection Machine 2 Computers in physics 6 6 643 646 1992 28 V M Fernandez N J Zabusky S Bhat D Silver and S Y Chen Visualization and feature extraction in isotropic navier stokes turbulence 29 J J Caban A Joshi and P Rheingans Texture based Feature Tracking for effective time varying data visualization 30 D Silver and X Wang Volume Tracking In R Yagel and G M Nielson editors IEEE Visualization 96 pages 157 164 1996 69 31 F Reinders F H Post and H J W Spoelder Attribute based feature tracking In Data Visualization 99 pages 63 72 Springer Verlag Wien 1999 32 G Ji H W Shen and R Wenger Volume tracking using higher dimensional isosurfacing In IEEE Visualization 03 pages 209 216 2003 33 F Y Tzeng and K L Ma Intelligent feature extraction and tracking for large scale 4D flow simulations In IEEE Supercomputing Conference pages 8 15 2005 34 R Samtaney D Silver N Zabusky J Cao Visualizing Features and Tracking Their Evolution IEEE Computer July 1994 p 27 7 20 27 35 M Gao Data Extraction and Abstraction in 3D Visualization Master thesis Rutgers University Elec amp Comp Eng Dept 1992 36 J
4. oe Places System GES en Y S WW Sat Mar 6 8 07 PM EISE EAR File Controls Options Windows Help a Window 1 SE aan SY poros MEE F S localhost ie 2 ANE B Ee x i oe at OC o no Loge Poly cycle 0003 pvorts poly cycle 0004 pvorts5 poly cycle 0005 Fvorts6 poly cycle 0006 ports Poly cycle 0007 Fvorts8 poly cycle 0008 Lotte Poly cycle 0009 vorts10 poly cycle 0010 H ReOpen Replace Overlay Regen ai a RE Active window Maintain limits _1 Replace Gs view data Auto update Active plots Hide Show Delete Cow Y K 1 Polydata TrackPoly var 1 Double Click here 5 Click here 4 Fill TrackPoly amp PolyData Attributes ral rr rr dr rr dr rr dr rr rl rr dd 0 20 40 60 NA 100 120 X 2 Select PolyData 3 Select TrackPoly from the list from the list user admin e Dee DEE EE ala avtFeatureTrackF I FeatureTrack Fil B admin miracle o Visit 1 12 0 DI Window 1 E Vorts File Brows I ou Palydata de atti E TrackPoly operat disk File Browser ja TE Figure 25 Visualizing the vorts1 poly file created by Feature Extraction amp Tracking module 1 Double click on the poly file as indicated in the figure 2 Click on plots and select PolyData plot from it 3 Click on Operators and select TrackPoly 4 Fill in the plot and
5. End Frame No P Threshold Smallest Obj vol Precision Raw Data E Database VtkDataset sea vtkDataset Pseudocolor gt Result Reader Plot format Operator vtk bov etc Generates colorma poly file attr file p file trak file uocd file traktable file per frame __ Specified by _ TrackPoly fa user Attributes weg e Path PolyData Attributes poly files Custom Colormap Path from step 1 ifany poly file vtkDataset TrackPoly vtkDataset PolyD ata Plot d Operator reader j j j 1 Fetches Update Visualizatio 1 Fetches curpoly txt n results curpoly txt 2 Extracts frame from Generates 2 Updates curpoly txt curpoly txt curpoly txt file with current 3 Fetches corresponding frame colormap file created in step 1 4 Forms vtkLookUpTable from colormap file 5 Apply to vtkDataset Figure 16 Two step process of Feature Tracking Step1 does feature extraction and generates files for step 2 Step 2 uses poly file created from step 1 and with TrackPoly operator and Polydata plot does the visualization and produces an output per frame and one trakTable file for the entire dataset The database reader loads a file provided the format is known to VisIt caches the metadata information about the file and passes on this information to the FeatureTrack operator The metadata information includes the number of nodes their connectivity scalar values at these nodes and sometime the value of normals at
6. which the program should read colormap information PolyData Attributes el Polydata plot attributes BEID TrackPoly operator attributes P Legend PolyFilePath home data vorts GENERATED_TRACK FILES P Lighting 4 Use min InitialcolorScheme Random v UserSelectedFile i Use mex Selected Scale Linear v Log y Skew Min S Make default _ Reset Max 1 Pos Dismiss Point size 0 05 Point Type 0 Skew factor 1 Opaciy Smoothing 0 4 Point size by var enabled Point size by var default r Point size pixels Poo Path JinomeldatavortsIGENERATED TRACK FILES SelectiveObjTracking 2 No v Yes Object Number to Track Opacity For Other Objects Leg q Make default Reset soy rel gel Figure 24 PolyData and TrackPoly Attributes screenshot e Path specifies the complete path to the location folder where poly files reside 74 e SelectiveObjTracking with this option user can choose to visualize only a specific group of object s and ignore the rest in a frame e ObjectNumbertoTrack Once the option for selective object tracking is selected a textbox is activated In this textbox user can enter the object numbers separated by a comma If nothing is entered in this textbox all objects would be tracked e OpacityForOtherObjects With this opacity slider user can determine how opaque the background objects can be The default value is 5 and it can be changed to any value between 0 and 100 via a slider
7. as dissipation e i A Take the remaining O in t as creation Figure 11 Overlap based Feature Tracking algorithm using Octree datastructure 24 The pseudocode for this algorithm is shown in Figure 14 Please refer to 3 4 30 for more information However this algorithm does not work with unstructured data and errors were noticed while tracking small objects 3 address the tracking problem with small objects An overlapping based feature tracking algorithm using a linked list data structure was developed for tracking unstructured grid datasets This algorithm can be extended to multiblock multiresolution and adaptive grid structures Features are extracted using a region growing algorithm 35 7 that generates an object list Each node in object list consists of object id attributes and all the nodes for that particular object Merging all the features of a frame and sorting them according to node ids generate a sorted node list The sorted node lists for two frames are compared to detect overlap Then best matching is performed on the overlap table to determine the class of interaction for each object in the frame 3 2 2 Parallel and Distributed Feature Tracking Algorithm The algorithms mentioned till now were incapable of handling large datasets efficiently Hence a distributed Feature tracking algorithm was developed In this algorithm features are merged using a complete merge 36 strategy that uses a binary swap algor
8. figure 13 defined and declared in avtPolydata C and avtPolydata h files The signature and behavior of these methods are described below 1 void avtPolyDataPlot SetAtts const AttributeGroup a 56 After the data is processed by the operator this method is called Here we can set the attributes for the plot In Polydata plot GetInfoAboutThisFile routine provides the information about the current file being handled by the plot After that we get the current polyfile and call setColorTable routine 2 void avtPolyDataPlot GetInfoAboutThisFile O This routines helps in finding the information about the poly files like the starting and ending frame number the number of steps dataset name This information is useful for verifying the current poly file that we are working on 3 vtkLookUpTable avtPolyDataPlot SetColorTable string This routine parsed the colormap file read the R G B and alpha values into different vector of doubles Creates a vtkLoookUp object and passes the R G B and alpha information to the vtkLookUpObject builds the lookuptable and applies it to vtkDataSetObject 5 5 2 Selective Feature Tracking Selective Feature Tracking requires a list of objects to be tracked The user provides this list as an input Hence it is one of the attributes to the PolyData plot User has an option of controlling the opacity for the objects that have to be removed to reduce visual clutter So another opacity slider is added as a
9. lt num node max gt Integrated content lt integrated content gt Sum of squared content values lt sum squared content value gt Volume lt volume gt Centroid lt x gt lt y gt lt z gt Moment Ixx lt xx gt lyy lt lyy gt Izz lt Izz gt Ixy lt lxy gt lyz lt lyz gt Izx lt Izx gt Sample attr file object 0 attributes Max position 63 000000 56 000000 127 000000 with value 12 139389 Node No 2087999 Min position 50 000000 66 000000 117 000000 with value 6 073445 Node No 1925426 Integrated content 52753 316406 Sum of squared content values 420312 062500 Volume 6857 Centroid 54 960667 66 905724 118 313576 Moment Ixx 81 173683 lyy 175 838699 Izz 76 425446 Ixy 105 836067 82 lyz 90 048668 Izx 46 851215 Object 1 attributes uocd This file is generated for each frame of the dataset It contains information about the frame particularly No of Objects Cell information etc This file is useful for volume rendering The format of the file is given below lt time gt lt num objects gt lt obj ID gt starts from 0 lt volume gt lt mass gt lt centroid X gt lt centroid Y gt lt centroid Z gt volume is of number of points lt Ixx gt lt lyy gt lt Izz gt lt Ixy gt lt lyz gt lt Izx gt lt point ID gt lt point X gt lt point Y gt lt point Z gt lt point value gt point ID starts from 0 lt point
10. o Examples of RPCs are GUI or CLI initiating state change in the viewer viewer causing the mdserver to perform an action such as opening a file viewer causing the engine to perform an action such as drawing a plot 31 CLI Local computer eee es me Remote computer 4 3 Workflow of Visit Consider a scenario where a user performs the following actions 1 Loads a data file 2 Choose a plot say contour plot 3 Choose a operator say slice operator 4 Click on Draw As a result of the above actions an AVT network Figure 8 is generated automatically by the VisIt We briefly mentioned about the network in the earlier chapter Here we see the network in detail We will see what methods are called when the above actions are performed Each user action corresponds to building some part of the network VisIt does not do any processing or visualization until the network is executed Till that time the Viewer sits with the information 32 1 Load Data File A user opens a file which causes the mdserver to open an avtFileFormat and get metadata information from the file This is the information about the data like the type of mesh scalar variables etc The actions are classified into two broad groups Meta data server actions 1 First MDServerConnection ReadMetaData method that is defined and declared in MDServerConnection C and MDServerConnection h is called e This method must first open a file so it
11. 25 datasets at a resolution of 35 41 23 4 Features were extracted from this dataset and tracking information provided visual cues on object evolution Isotropic Turbulent Decay Simulation Feature Tracking was applied to LES simulation of the decay of istrophic turbulence in a box in a compressible flow using unstructured tetrahedral This simulation dataset having 500 frames or timesteps showed that the number of objects changes with as the isotrophic turbulence decays 4 Autoignition datasets Basic feature tracking algorithms can be useful as an analysis tools for combustion datasets by application to a dataset modeling autoignition 28 In 28 Features defined as areas of high intermediate concentrations were examined to explore the initial phases in the autoigniton process Turbulence flows Feature tracking can be useful in identifying and temporally tracking hair pin packets and their wall signatures in direct numerical simulation data of turbulent boundary layers 29 In this work visualization algorithms are validated against the statistical analysis And they demonstrate that the average geometric packet is representative of strong statistical ones Also they presented for the canonical case of an isolated hair pin packet convecting in channel flow and for fully turbulent boundary layers 26 3 4 Software Implementations of Feature Tracking Algorithms 3 4 1 AVS Express implementation The Ostrk2 0 package is a stand alone
12. Feature Tracking tracks these features over a period of time Tracking features play an important role in the studying the evolution of different physical phenomenon and scientists can build predictive models based on these analysis Feature tracking can be very useful in analysis of natural phenomena like hurricanes and development of a prediction system to minimize the damage caused by these phenomena Figure 11 shows results from application of Feature Tracking technique on hurricane data from Hurricane Bonnie 29 In Figure 11 a one feature was tracked for 30 timesteps b three independent features were tracked and c a number of independent features were tracked over time From these results it was possible to see that most of the features under consideration followed a clear pattern such as moving counter clockwise and inwards It was also possible to see that features closer to hurricane s center moved faster than the features farther from the hurricane s eye Both of these findings were extremely important in terms of the analysis and interpretation of hurricane data a b Figure 8 a Results of tracking a hurricane feature within 30 timesteps b Path followed by three independent features over the same period of time C Resulting path after tracking a number of independent features over time Image source 29 20 3 1 Feature Tracking Techniques Over the years researchers have proposed several different techni
13. ID gt lt point X gt lt point Y gt lt point Z gt lt point value gt lt point ID gt lt point X gt lt point Y gt lt point Z gt lt point value gt lt objID gt Sample uocd file 10 000000 38 12637 83432 195312 68 521065 97 046165 99 475235 83 234 301193 51 264282 245 409286 85 529694 79 158249 222 195129 49 1798593 65 000000 99 000000 109 000000 10 520865 trak This file is also generated for each frame of the dataset This is like a high level summary of the objects in the frame It contains number of objects and number of nodes to be used for memory allocation for tracking with successive timestep The format of this file is given below lt file base name gt lt time gt lt mass gt lt volume gt lt centroid X gt lt centroid Y gt lt centroid Z gt lt file base name gt lt time gt lt mass gt lt volume gt lt centroid X gt lt centroid Y gt lt centroid Z gt Sample trak file home admin data GENERATEDTRACKFILES vorts10 10 000000 83432 195312 1263768 521065 97 046165 99 475235 home admin data GENERATEDTRACKFILES vorts10 10 000000 42605 246094 631662 313812 117 506172 39 228920 home admin data GENERATEDTRACKFILES vorts10 10 000000 40073 558594 577162 039474 46 921017 4 576792 trakTable There is only one trakFile for the entire dataset This file is not created if there is only one frame in the dataset otherwise this file is created from second frame and with each new
14. In the terminology of AVS Express uniform rectilinear and irregular fields all exist as structured meshes They consist of nodes arrayed on a one two or three dimensional hexagonal grid This is all the information that is required to know how the nodes are organized and connected in space 90 Structured 2D Mesh Structured 3D Mesh Image source http sciviz aict ualberta ca index php page data amp sub_page structure Unstructured Mesh The unstructured mesh is the only type which must have the connectivity information provided to the module as an input In unstructured meshes where cell data is unstructured nodes exist in space but as part of a higher organization called a Cell The nodes are connected together in a certain order to form point line triangle quadrilateral tetrahedral pyramid prism or hexahedral cells These cells collectively form the field Unstructured 2D Mesh v1 Unstructured 2D Mesh v2 Unstructured 2D Mesh v3 Image source http sciviz aict ualberta ca index php page data amp sub_page structure The language used to describe this mesh varies between the packages as follows Mesh Type AVS Vtk or VisIt Uniform Uniform Mesh Structured Point Rectilinear Rectilinear Mesh Rectilinear Grid Irregular IrregularMesh StructuredGrid i sti itstsC Y Structured Structured Mesh Structured Grid Unstructured Unstructured Mesh Unstructured Grid
15. Plugins Plugin Mod Class Mod Added Method Functionality Extractian amp Feature raciona i m avtFeatureTrackFilter BeginFeatureTrack Tracking Plugins Modified for creating colormaps ReadCurrentFrameTr ackInfo find_merge_color Tokenize Feature Correct Tracking Visualization Plugins TrakTable based coloring GetMesh Poly f avtPolyFileFormat p Modified to create curpoly txt ExecuteData Track eee ES DE kg avtTrackPolyFilter p Modified it for Poly udatingcurpoly txt Visualization Group Plugins Geesen ParseColormapfile avtPolyData GetInfoForThisFile SetColorTable CreateOpacityTable PolyData S A GetinfoAboutThisFile Selective startsWith Feature toString Tracking Tokenize ReadCurrentFrameTrackInfo SeperateCurrentAndNextFram eObjects PopulateNextFrameObjects Picking UpdateObjectPicked L Objects by PickedObject Mouse Clicks Figure 17 A Tree list view of methods added in corresponding classes for achieving the functionality shown on extreme right 52 Figure 19 is a tree structure describing all new methods added for achieving the target Feature Tracking amp Visualization functionalities The methods are grouped according to the functionality and indicate the classes that need to be modified in respective plugins to accommodate new methods In the figure only the names of the methods
16. Results of Feature Tracking on 2D Dataset The results from visualization of these datasets have been very similar to the ones produced from AVS However we could not make comprehensive performance calculation as both were running on machines of different configurations Newer versions of VisIt were not compatible on old Red Hat machines in the lab as result we had trouble compiling VisIt on Red Hat Linux 7 0 AVS could not be installed on newer machines because of license issues Hence we could not run AVS and VisIt together to compare the running times of both softwares However previous work has tabulated this and Vislt performs better than AVS 23 67 7 Conclusion Feature Tracking is a really powerful tool for studying the evolution and tracking of features in time varying datasets VisIt s rich feature set makes development of Feature Tracking plugins in VisIt much more easier and simpler Feature tracking adds a lot of value to VisIt and could be useful in studying any kind of 3D datasets Feature tracking of 2D datasets can also be performed with this plugin increasing it usability to a wide range of users With size of datasets growing bigger and bigger VisIt s parallel processing feature would be a great tool and Vizlab s stand alone distributed mode Feature Tracking module would prove to be more than handy The next generation of Feature Tracking is aiming to be smarter and provide a lot more functionality than before like
17. VisIt s bin folder is added to system path To create new plugins using these tools one must follow the following steps 1 2 3 4 Create a directory by the plugin name The location of this folder depends on the type of the plugin If the plugin is operator it should go under lt visithomefolder src operators gt if its a plot it should be in lt visithomefolder src plots gt and if it s a database plugin then it should be at lt visithomefolder src database gt Change to the location of the directory and run xmledit by following commands cd lt visithomefolder src lt plugintype gt lt pluginname gt gt xmledit A window similar to Figure 11 would appear on the screen An untitled xml file is opened for the user to fill in the information necessary for creating a plugin The information includes the name of the plugin the type of the plugin attributes and so on refer to VisIt Developer s Manual for more details Attributes are the parameters that allow users to interact with the plugin For example opacity for a plot can be changed via a slider A plugin can have one or more than one attributes After the attributes are selected their description entered and all the information provided the file is saved as an xml file by the same name as that of the plugin So if the plugin is TrackPoly then the xml file should be saved as TrackPoly xml in TrackPoly folder Run xml2plugin This will automatically create a f
18. feature tracking software developed in C C on the AVS Express 6 2 platform by Vizlab 17 This software is implementation for linked list based overlapping Feature Tracking technique This software works not only with unstructured datasets but all other types of datasets too The main features of this software package are summarized below e Feature Extraction The input dataset is segmented into its features as threshold specified by the user connected components The user choose a percentage threshold and the actual value for this percentage threshold is actual thresh p max node value min node value 100 where p is the percentage threshold selected This calculation is performed on a per frame timestep basis e Feature Tracking The life cycle of all extracted features is tracked over the number of time steps of the dataset specified recording all events that may occur during an objects life cycle specifically merging splitting continuation dissolution or creation Enhanced Surface Animation Users can view an iso surface visualization of all time steps with color coding added to highlight feature events For example suppose a feature A in time step 1 splits into features B and C in time step 2 Then both features B and C will receive the same color as A e Surface Isolation Animation The interface in Ostrk2 0 allows you to select a particular feature from the surface animation window last time step only and
19. fields that allow new fields to be calculated using existing fields For example if a dataset contains a velocity field it is possible to define a new field that is the velocity magnitude e Supports multiple mesh type VisIt provides support for a wide range of computational meshes including two and three dimensional point rectilinear curvilinear and unstructured meshes In addition VisIt supports structured AMR meshes and CSG meshes e Powerful full featured Graphical User Interface GUI VisIt s graphical user interface allows novice users to quickly get started visualizing their data as well as allowing power users access to advanced features VisIt automatically creates time based animations from data sets that contain multiple time steps e Parallel and distributed architecture for visualizing tera scale data VisIt employs a distributed and parallel architecture in order to handle extremely large data sets interactively VisIt s rendering and data processing capabilities are split into viewer and engine components that may be distributed across multiple machines 16 e Interfaces with C Java and Python VisIt also supports C Python and Java interfaces The C and Java interfaces make it possible to provide alternate user interfaces for VisIt or allow existing C or Java applications to add visualization support The Python scripting interface gives users the ability to batch process data using a powerful scripting lan
20. for each frame or timestep in the dataset It contains node co ordinates of the polygon vertices and their connectivity information for iso surface visualization A poly file looks like lt red gt lt green gt lt blue gt lt numnodes gt lt x0 gt lt y0 gt lt z0 gt lt xl gt lt yl gt lt z1 gt lt x2 gt lt y2 gt lt z2 gt lt x3 gt lt y3 gt lt z3 gt lt x4 gt lt y4 gt lt z4 gt lt numof connections gt 3 lt vertex D gt lt vertexlI D gt lt vertexl D gt 3 lt vertex D gt lt vertexlI D gt lt vertexl D gt 3 lt vertex D gt lt vertexlI D gt lt vertexl D gt 0 signifies end of information about a object in the frame lt red gt lt green gt lt blue gt continues for next objects followed by 0 and so on 80 Sample poly file 0158 96 6774 67 234009 102 000000 112 000000 68 000000 102 000000 111 367813 68 000000 101 106598 112 000000 13488 3132 3465 3798 0 102550 attr Generated for each frame of the dataset this file contains information about each object in the frame like its centroid moments max node value etc It looks like this Object lt obj ID gt attributes Max position lt max X gt lt max Y gt lt max Z gt with value lt max node value gt Node No lt num node min gt 81 Min position lt min X gt lt min Y gt lt min Z gt with value lt min node value gt Node No
21. object s provided as input and writes them to a file called OpacityTable txt see appendix I for file format The helper methods for creation of opacity table are as follows 2 void ReadInfoForThisFile This methods help to parse the CurrentFile txt file see appendix for file format 3 bool startswith string string Checks whether a given string a sentence starts with the specified pattern and return true if it matches else returns false 4 string toString int This method converts a given integer type value into a string type value 5 vector lt string gt Tokenize const string amp Given a constant reference to sentence i e an address location have a string which cannot be altered this method breaks into words separated by comma and returns the results as a vector of those string tokens 6 void ReadCurrentFrameTrackInfo const string amp const int amp map lt string string gt amp This method reads trakTable for a particular frame number and creates a map for the objects in that frame So every object in the current frame is mapped to an object or group of objects in next frame The map data structure allows finding of corresponding objects easily 24 7 void SeperateCurrentAndNextFrameObjects const vector lt string gt vector lt string gt amp vector lt string gt amp Suppose an entry in trakTable refer appendix I for file format file is like this 58 14 7 1 3 Whatever is to
22. the Clear menu in the Main Window s Window menu The dimension of the plots in the visualization does not matter when using pick mode Both 2D and 3D plots can be picked for values However when using pick mode with 3D plots only the surface of the plots can be picked for values If you want to obtain interior values then you should use one of the Pick queries or apply operators that expose the interiors of 3D plots before using pick The Pick Window mainly consists of a group of tabs each of which displays the values from a pick point The tab labels A B C etc corresponds to the pick point label in the visualization window Since there are a fixed number of tabs in the Pick Window tabs are recycled as the numbers of pick points increases When a pick point is added the next available tab which is usually the tab to the right of the last unused tab is populated with the pick information If the rightmost tab already contains pick information the leftmost tab is recycled and the process repeats The information displayed in each tab consists of the database name and timestep the coordinates of the pick point the zone cell that 92 contains the pick point the nodes that make up the cell containing the pick point and the picked variables The rest of the Pick Window is devoted to setting options that format the pick output The window on the right hand side of figure 20 is pick Window AVS to vtk converter VisIt does not understand
23. the Hd file format of AVS Hence we wrote a program to convert the fld files into vtk format The source or the binary files for Linux and Windows platform can be downloaded from http miracle rutgers edu download html Sample fld file AVS field file ndim 2 number of computational dimensions in the field dim1 7 Dimension of axis 1 dim2 4 Dimension of axis 2 dim3 1 Dimension of axis 3 nspace 2 number of physical coordinates at each point veclen 1 number of vector components at each point data float the data in the file is of type byte integer float or double field uniform uniform rectilinear irregular min_ext 0 0 lower bounds of x y and z axes max_ext 6 3 upper bounds of x y and z axes label radius label the field components unit arbitrary units for the field components For a uniform field read in coords of points with min and max values coord 1 file unif_2D dat filetype ascii skip 11 offset 0 stride 3 coord 2 file unif_2D dat filetype ascii skip 11 offset 1 stride 3 coord 3 file unif_2D dat filetype ascii skip 11 offset 2 stride 3 93 Read variables in like this read data values in 4th column Don t need the coordinate data in the first three columns variable 1 file unif_2D dat filetype ascii skip 13 offset 3 stride 4 Sample vtk file vtk DataFile Version 1 0 Structured Points 3D Dataset ASCII DATASET STRUCTURED_POINTS DIMENSIONS 7 4 1 ORIGIN 000 SPACING111 PO
24. the frame A new object is formed in frame 7 and keep on growing Also in Frame7 two objects that were continuing as separate objects till now combine to form a bigger object in frame 8 one of the objects disappear 2 DNS Dataset This dataset consisting of 27 frames is used in the study of hairpin packets of DNS data for Mach 3 turbulent boundary layer Each frame of the dataset has a resolution of 384 256 69 and each point stores a scalar value float type We select a feature of interest in the first frame object 45 and see its evolution for next few frames The results for selective tracking of this particular object are given in Figure 25 For getting this dataset to work in VisIt we had to write a conversion program for converting the data to a format known to Vislt We converted the files from Hd format used by AVS to vtk for VisIt and the details of this program are give in Appendix II 64 DB reconbt1 Gol Cvcle 1 Bee 0 0 1 0 2 0 30 X AS 5 0 6 0 DB reconbt2 poly Cycle 2 Time 0 0 1 0 2 0 3 0 X AS 5 0 6 0 DB reconbt3 poly Cycle 3 Time 1 0 0 1 0 2 0 3 0 X AS 5 0 6 0 Figure 20 Selective Feature Tracking for Object 45 on DNS dataset 3 Surfactant molecular Dataset Modeling and Simulation Department P amp G This dataset of 101 frames is named volmap and resolution of each frame is 47 47 47 points A scalar value float type is stored at each of those points We use
25. the left of delimiter 1 belongs to current frame and whatever is to the right are objects from next frame This method separates the entries in the TrakTable file into current frame and next frame objects 8 void PopulateNextFrameObjects mapsstring string gt vector lt string gt amp vector lt string gt El This method creates a vector by referring to the map data structure and pushing all the objects for the next frame into a vector It s possible to have duplicate entries in the vector we remove the duplicates and ensure that all the elements are unique 5 5 3 Picking Objects by Mouse Click VisIt uses Observer also called subscriber publisher design pattern in which a Subject maintains a list of its dependents called observers and notifies them automatically of any state change usually by calling one of their methods 25 VisIt provides support for picking object in Zone Pick and Node Pick modes see Appendix II for more details and these modes subscribe to Pick attributes Pick attributes keep a track of current mouse coordinates and finds out whether the click is on area occupied by an object or empty space Whenever there is a change in Pick Attribute pick modes are notified of it via UpdateWindow method This method is defined for every plot in Qvis lt Plotname gt Window class If the plot has to be made aware of changes in Pick Attributes it needs to subscribe to Pick Attributes Plots have certain subjects that it kee
26. these nodes Now the 50 FeatureTrack operator is ready to segment the data extract the features and write the information about the extracted features to the files Feature Track operator needs to know some more information about the dataset like the starting and ending frame numbers the percentage threshold for segmenting objects the location of the files and a few other things for more details look at FeatureTrack Attributes chapter 6 These parameters are called Attributes and they are a mechanism for providing user input to the plugins These attributes have to be provided each time to get intended results These attribute values are highly dependent on the datasets and kind of data they represent Hence providing a default value does not make sense FeatureTrack operator generates all the necessary files and passes on the vtkDataSet object to the plot One may ask our aim was to generate these file and FeatureTrack operator achieves this then why do we need to use a plot at all Isn t it redundant VisIt does not allow the application of an operator to the data directly without selecting a plot hence we had to choose a plot Since the plot has no useful work to perform and has no bearing on results we just choose any plot we decided to use pseudocolor plot as dummy The output on the screen varies from one to two objects or a blank screen This is fine as along the intermediate file are generated Once the files are generated from step
27. to produce piece invariant results Ghost levels are points cells shared between processes and are used by algorithms which require neighborhood information ParaView supports both distributed rendering where the results are rendered on each node and composited later using the depth buffer local rendering where the resulting polygons are collected on one node and rendered locally and a combination of both for example the level of detail models can be rendered locally whereas the full model is rendered in a distributed manner This provides scalable rendering for large data without sacrificing performance when working with smaller data Scripting and extensibility ParaView is fully scriptable using simple but powerful Python language Additional modules can be added by either writing an XML description of the interface or by writing C classes The XML interface allows users developers to add their own VTK filters to ParaView without writing any special code and or re compiling 14 From data to Pictures in ParaView The procedure for converting data to picture is ParaView is also similar to that of AVS First the data has to be read into ParaView Since ParaView is a subset of VTK ParaView supports most of the file formats supported by VTK Incase ParaView could not find the reader associated with a particular file format then additional reader module has to be written to read the data in ParaView Once the data is read in ParaView th
28. when the user executes the network CurrentFile txt is written to the folder where the poly files reside TrackPoly operator and Polydata plot get the location of the folder from their attributes that user enters Whenever VisIt s opens a new file CurrentFile txt gets over written VisIt loads the new data loaded and issues an RPC to the operator and plot to work on this new data Hence we can be sure that operator and plot always have the updated information as RPC is issued only after CurrentFile txt is overwritten G pions Sen ste BB OH S suns s ssm H 11 SO Ele Conois Options Windows Hep O aaa WEE KZ BABO All the Files are grouped as a Database All this eT lf information is part of e AW e QvisFilePanel methods Active window Maiti its Replace pios 1 i wenig 1 Auto update Sep zegsl zs om CEE ATA a CIT A e 61 Figure 18 Grouping of all the files as one database This is done automatically by VisIt and is part of QvisFilePanel class 62 6 Results Feature tracking helps in studying the evolution of pattern We were mainly interested in tracking the following patterns e Continuation of objects e Merging to form new objects e Bifurcation e Dissipation e Creation of new objects We ran Feature Tracking in VisIt on the following datasets and tracked the above patterns in all of those datasets The functionality of selective feature tracking and pickin
29. 1 it s the task of visualization plugins to display the results on the screen The information about the extracted objects is in poly files and these files are an input to step 2 The format of poly files is unknown to VisIt hence Poly database reader was written for this purpose 23 During the loading of the poly files the plugin generates a curpoly txt text file This file has information about the file name and number of objects in the file and current frame number which is extracted from file name Now the vtkDataSet object created by reading poly files is passed to TrackPoly Operator The operator does nothing to vtkDataSet object but it updates the frame number in curpoly txt and passes the vtkDataSet object to the plot The plot opens the curpoly txt file reads the current frame number and fetches the corresponding colormap file for the frame these files are created in step1 From the colormap information a vtkLookUpTable objects is created This is mechanism by which user can specify their own colors to VisIt and VisIt renders accordingly vtkLookUpTable is an array of R G B and alpha values Reading the colormap file fills these values and the information is passed to vtkDataSet 51 object After calling appropriate methods colormap is applied to the objects and the results are rendered on the screen The process then repeats for other files or frames in the database 5 5 Modification to Feature Tracking amp Visualization
30. 360 170601 170106 040437 127277 124446 136631 041462 116321 0266400 020243 005602 004146 121574 124651 006634 071331 102070 0266420 157504 160307 166330 074251 024520 114433 167273 030635 0266440 133614 106171 144160 010652 007365 026416 160716 100413 0266460 026630 007210 000630 121224 076033 140764 000737 003276 0266500 114060 042647 104475 110537 066716 104754 075447 112254 0266520 030374 144251 077734 015157 002513 173526 035531 150003 0266540 146207 015135 024446 130101 072457 040764 165513 156412 0266560 166410 067251 156160 106406 136770 030516 064740 022032 0266600 142166 123707 175121 071170 076357 037233 031136 015232 0266620 075074 016744 044055 102230 110063 033350 052765 172463 Figure 1 Visualization of array of number as an image Image source Ch 1 Page 12 http www cmake org Wiki images 6 65 ParaViewTutorial36 pdf In addition time varying simulations are common in many scientific domains these are used in the study of evolution of phenomena or features Traditionally visualization of 3D time varying datasets is done using animation Le each frame or time step is visualized using iso surfacing or volume rendering and then the various time steps are run in sequence to decipher any visual patterns However for datasets with continuously evolving features like cloud formations it is difficult to follow 6 and see patterns in 3D What is required is a technique to isolate and analyze certain regions of interest also c
31. ABSTRACT OF THE THESIS Feature Tracking amp Visualization in Vislt by Naveen Atmakuri Thesis Director Professor Deborah Silver The study and analysis of huge experimental or simulation datasets in the field of science and engineering pose a great challenge to the scientists Analysis of these experimental and simulation datasets is crucial for the process of modeling and hypothesis building Since these complex simulations generate data varying over a period of time scientist need to glean large quantities of time varying data to formulate hypothesis or understand the underlying physical phenomenon This is where visualization tools can assist scientists in their quest for analysis and understanding of scientific data Feature Tracking developed at Visualization amp Graphics Lab Vizlab Rutgers University is one such visualization tool Feature tracking is an automated process to isolate and analyze certain regions or objects of interest called features and to highlight their underlying physical processes in time varying 3D datasets In this thesis we present a methodology and documentation on how to port Feature Tracking into VisIt VisIt is a freely available open source visualization software having a rich feature set that can help scientists visualize their data and provide a powerful data analysis tool VisIt can successfully handle massive data quantities in the range of tera scale The technology covered by this thesis i
32. Chen Y Kusurkar D Silver Distributed Feature Extraction and Tracking SPIE Visualization and Data Analysis Symposium January 2002 37 C Hansen T Udeshi S Parker P Shirley Parallel methods for isosurface visualization extended abstract Ninth SIAM Conference on Parallel Processing for Scientific Computing 1999 38 Pinakin s thesis 70 Appendix I Installation of Feature Tracking amp Visualization plugins in VisIt After installing the VisIt on Linux VisIt creates 3 main directories under the src directory namely operators database and plots Any new plugins should be created under one of these folders based on its type Automatic Installation The source code for all plugins of Feature Tracking can be downloaded as tar file from http miracle rutgers edu If VisIt is not installed on the machine follow the instruction in 26 to install VisIt All these instructions are for Visit Version 1 12 0 and might be somewhat different for other versions The steps are as follow 1 Download Plugins tar and InstallAllPlugins files under your lt Yourvisithomefolderpath gt 2 Run the script file under your visit home folder lt Yourvisithomefolderpath gt as follows lt Yourvisithomefolderpath gt nstallAllPlugins If the above steps does not yield any errors then the installations of all plugins is complete On opening VisIt new plugins should be visible under either plots or operators Manual Ins
33. Dismiss Figure 23 FeatureTrack operator attributes screenshot Visualization After poly files are created in the first step they need to be visualized This is done as follows 1 Select a poly file by double clicking on it or selecting the auto generated poly database 2 Select PolyData plot from the list of plots 3 Select TrackPoly operator 4 Fill in the Operator attributes and Plot attributes Once you have entered the attributes click on apply in both plot and operator 5 Click on Draw to get the results see Figure 19 73 TrackPoly Attributes e PolyFilePath The path location where the poly files reside InitialColorScheme User can either select Random or UserSelectedFile If Random option is selected then TrackPloy operator will randomly assign colors to all the objects in the first frame or timestep In subsequent frames each object will be assigned a color depending on the tracking information If the user selects the UserSelectedFile then the SelectedFile attribute is enabled and user can give the complete path of the user defined colormap for the first frame TrackPloy operator will assign colors to all the objects in the first frame based on this colormap file Again in subsequent frames each object will be assigned a color depending on the tracking information e SelectedFile In case user descides to provide their own colormap This is the location from
34. INT_DATA 28 SCALARS scalars int LOOKUP_TABLE default 1000000000000 0 0000000000000 1 As we can most of the data in these files is similar it easy to parse the Hd file and then create appropriate vtk format for the part parsed Only difference comes in coordinates AVS files can have coordinates in a separate binary file vtk does not allow this and they have to be presesnt in the same file under a lookup table The data here can be in ASCII or Binary mode Visualization data is such that there are lot of 0 s in between hence the size of the fize doesn t increase much and are smaller in size compared to binary files 94 2D to 3D converter As our Feature Tracking does not work on 2D data we have to fake it as 3D data by adding an additional layer of points equal to the product X amp Y coordinate points The value given to the points in the new layer is the minimum value in the dataset We just added one additional layer but more layer can be added if needed The parsing of fld files and creating vtk is same as the one discussed in previous section After this an additional section has to be added to process the vtk file The sample vtk file given in previous shows is a 2D file even in 2D file the value of Z coordinates is shown as 1 as all we have to do is update it toa value of 2 and add more points to LOOKUP_TABLE Once this is done our 2D Dataset can be used for Feature Tracking in VisIt environment The sour
35. The first phase detects the overlaps among features and limits the number of candidates to be matched in second phase This phase has three steps Segment the dataset into objects and background and store the nodes for each object in the object list Merge the object lists and sort in ascending order of node ids Compare the two sorted lists from t to ti to detect the overlap and store these results in overlap table 2 Best Matching test this phase finds the correlation between different features Feature Tracking For two consecutive time steps t and t Extract all the features from the two databases and store each feature in its own octree i 1 Construct the octree forests F U O and F U OF pet dei Lee O as a template for matching For each feature O F merge it into the octree forest F to Identify all the overlapping regions of O in ti Store this in a list called OverLap GE For each feature O in F Determine bifurcation and continuation For all combinations of features in OverLap O IL Compute O _U Overlapo If the lowest difference is below the tolerance Mark O as bifurcating into the object and remove them all from the search space i Next O E Else Determine Amalgamation SC A 41 pi For each remaining feature in O merge it into the octree forest F and test for amalgamation This is the same as bifurcation with the inputs Take the remaining O in t
36. ability in Vislt e Implement the functionality to selective track objects e And pick the objects to be tracked by Mouse clicks e Design a flexible intuitive easy to use and solid framework that lays a strong foundation for next generation Feature Tracking Algorithm e Document the design decisions and workflow of Visit which can be useful to other users for adding new capabilities e Organize the Feature Tracking code and collect all datasets at once place Apart from many wonderful features of VisIt discussed earlier there is one strong technical reason for implementing Feature Tracking amp Visualization in VisIt The architecture of VisIt is such that it aides in separation of the two main modules for Feature Tracking amp Visualization process These modules are 41 e Feature extraction and Tracking module e Visualization module Incase of AVS features are extracted for one frame tracking information updated and the results are rendered on the screen Then AVS moves to next frame and repeats the process of extracting features updating tracking information and rendering the result But VisIt works in a different way First features one by one from all frames simultaneously updating the trakTable After performing this function then only the results are rendered on the screen How is this separation of modules advantageous Suppose a scientist is analyzing some experimental simulation dataset After 25 frames scientist sees s
37. alities are available to users free of cost VisIt is one such open source visualization software that is freely available and has a rich feature set VisIt supports quantitative and qualitative visualization it has powerful user interface and architecture to support massive datasets in the order of tera scale VisIt allows development of custom plugins to add new features and functionality Feature Tracking capability can be introduced in VisIt as plugins and this would allow VisIt user to study evolving patterns in time varying datasets Previous work on porting Feature Tracking to VisIt 23 extracted the features from a time varying scalar dataset However the visualization plugins were found to be incomplete If an object is moving across a dataset changing its shape and size the algorithm should be able to identify this phenomenon 7 and assign the same color to object in other frames until they disappear Even if the object splits into smaller objects all those smaller objects should continue to have same color until they disappear Since this kind of behavior is depicted in trakTable we use trakTable to assign colors and present an algorithm to achieve this task Our implementation gives the scientist or a user an option to selectively track only few features or objects of interest amongst all the extracted features These features can be selected by mouse clicks Like most open source softwares VisIt too lacks proper design documenta
38. alled features and highlight their underlying physical processes 1 2 For example it is usually important to know what regions are evolving whether they merge with other regions and how their volume may change over time Therefore region based approaches such as feature tracking and feature quantification are needed Moreover most of the standard visualization methods cannot give a quantitative description of the evolving phenomena such as the centroid of a merged region or the value of its moments An automated procedure to track features and to detect particular stages or events in their evolution can help scientists concentrate on regions and phenomena of interest Color connected iso surface based on computed quantification provide visual cues about the events or particular stages of interest This effectively reduces the amount of data to focus on by curtailing the visual clutter Another important application of feature tracking is in data mining By building a database of features the results over multiple simulations can be used for event matching In previous work 3 4 5 the Vizlab had pioneered the use of feature tracking to effectively visualize time varying datasets Feature Tracking was implemented as a plugin on AVS a proprietary visualization software Thus license costs limit the usability and extendibility over a larger user base Many open source visualization software packages that contain lot more features and function
39. are given and not their signatures These methods and their signature are discussed later on in this section according to the functionality for which they were added modified 5 5 1 Traktable based coloring From the workflow of Feature Tracking algorithm figure 18 FeatureTrack operator seems to be the best place for generation of colormaps and PolyData plot for the application of the colormap to the data Implementing this functionality involved changes in all the 4 plugins 5 5 1 1 New methods in FeatureTrack Operator for Colormap generation The colormaps are generated by modifying and adding methods in avtFeatureTrackFilter class see figure 19 defined and declared in avtFeatureTrackFilter C and avtFeatureTrackFilter h files The signature of these methods and their behavior are given below 1 int avtFeatureTrackFilter BeginFeatureTrack char int amp The following steps were added to this method at the end of the existing definition 1 Get the current frame number 2 Ifitis the first frame Open a new file called colormap1 txt Get number of objects in current frame Run a loop from 0 till number of objects in the frame for each iteration e generate three random integer values use rand and store them as R G and B values Let alpha be 1 e Write down the all the three R G B and alpha to the colormap1 txt file End the loop and close the file 53 3 Ifit s not the first frame then open a new c
40. ase Reader avtFeatureTrackFilter C avtPolyDataPlot avtTrackPolyFilter avtPolyFileFormat h C h C h C Ftrack FeatureTrack QvisPolyDataPlotW C h indow Ch Ftrack FeatureTrackUtil h Interface InterfaceUtil h ObjSegment cellinfo C h ObjSegment nodeinfo C h ObjSegment objectinfo C h ObjSegment input h ObjSegment ObjectSegmen t C ObjSegment ObjSegmentU til h ObjSegment stobject1 h ObjSegment stRGB h ObjSegment Util Ch Table 3 List of all files created modified indicates the auto generated files which were modified 5 4 Feature Tracking Visualization Workflow in Vislt Feature tracking is a two step process Figure 18 The first step extracts and tracks the features while the second step visualizes the features Step2 is dependent on Step1 as some of the files like poly are generated in Step1 are input to Step2 Step 1 has to be executed atleast once to generate all the files Once the files are there Step2 can be performed as many times as user wants This separation of modules gives VisIt an advantage over AVS Step 1 creates the files as mentioned in figure 13 5 files Step 1 49 Step 2 FeatureTrack Attributes age Path SE Specified by Start Frame No om i user
41. ata e eighty five file format readers over fifty queries ways to extract quantitative information e over one hundred expressions ways to create derived quantities Further a plugin capability allows for dynamic incorporation of new plot operator and database modules These plugins can be partially code generated even including automatic generation of Qt and Python user interfaces 4 1 High level design of Vislt VisIt is composed of multiple separate processes which are sometimes called as components 19 They are listed in Table 1 Name Overview Viewer Two primary purposes First it centralizes all of VisIt s state When the state changes it notifies the other components of the state changes Second the viewer is responsible for managing visualization windows which often includes doing rendering in those windows Gui Provides a graphical user interface to control VisIt Cli Provides a command line user interface to control VisIt 30 Vel Launches jobs on remote machines The VCL sits idle on remote machines communicating with the viewer and waiting for requests for jobs to launch When these jobs come up it launches them The purpose of this module is to spare the user from having to issue passwords multiple times Mdserver The mdserver browses remote file systems meaning it produces listings of the contents of directories It also opens files in a lightweight way to get
42. butes trak and table file local object table for its local region The code in this directory is compiled using a MPI compile script finalmerge directory This part of code includes methods to read in the table files generated during extraction step to generate a global object table Also poly and oct files are updated in this step ftrack directory here distributed tracking is implemented by partial merge strategy Again MPI compile script is used to compile the code in this directory score directory best match is calculated and features across the frames are correlated Each of these directories have to be compiled separately to get the final results The order of compilation should not be changed as the output of a particular step is needed by another step Since the process of compiling these smaller programs manually was becoming cumbersome a perl script was written to automate this process of distributed feature tracking More information about the script its implementation and the algorithm is given in Pinakin s thesis 28 3 4 3 Visit Implementation Feature Tracking of time varying scalar datasets in VisIt environment 23 describes porting of Feature Tracking in VisIt This implementation successfully extracted the features from datasets and tracked those features in VisIt environment Information about the extracted features were store in poly attr uocd and trak files while the tracking information was
43. by merging or continue as 42 before Since the object continues to exist although in a different form in rest of the frames it should get the same color as in first frame If a new object is formed in a particular frame then it should be assigned colors that do not correspond to any of the existing objects Tracking information is stored in trakTable file Hence a color scheme based on trakTable would produce images similar to those on the right side of figure 14 When the objects following such a color scheme it becomes very easy for the user to track these objects visually In Figure 14 we can easily identify that Object A continues in 2 frames and splits in the last frame while object B moves in the all the frames of the dataset with a very little change in its size and location Such kind of behavior is useful in analysis and process modeling 5 2 2 Selective Feature Tracking Sometimes we just want to visualize and track only those objects that are of interest to us We should therefore have a mechanism to specify a list of objects to be tracked This list when passed to the program should render only those objects in the list and remove other objects from the screen Instead of actually removing the objects we can adjust the transparency value to make them almost transparent and reduce visual clutter Users should have the ability to control the opacity value for both the object to be tracked and the objects that are to be made transpa
44. calls MDServerConnection GetDatabase This method uses an avtDatabaseFactory to instantiate an avtDatabase The DB factory iterates over viable plugin types 1 For each viable plugin type the file format is instantiated 2 avtDatabase GetMetaData is called This forces the plugin to do some work to see if the file is really of the format s type 1 GetMetaData ultimately calls PopulateDatabaseMetaData GetCycles GetTimes 2 No calls will be made to GetMesh GetVar e The resulting avtDatabase is asked to create meta data for the file This is a no op since the meta data was read when opening the file and that meta data was cached 2 Later SIL information is requested of the database This is done in MDServerConnection ReadSIL using avtDatabase GetSIL 1 avtDatabase PopulateSIL is called 33 2 avtSILGenerator populates the SIL entirely from the meta data Engine actions The first of the engine actions it to call the method RPCExecutor lt OpenDatabaseRPC gt Execute defined in Executors h 1 The appropriate plugin type is known from the mdserver which is open and it is loaded 2 This calls NetworkManager GetDBFromCache which does the following 1 The file is opened using the database factory 2 avtDatabase GetMetaData is called 3 avtDatabase GetSIL is called 4 The database is registered with the load balancer 2 Adding a plot or operator as the user makes the plot and adds operator the engine responds by cons
45. ce code can be downloaded from http miracle rutgers edu download html
46. ciated with ObjectNumToTrack get the object number of the selected object and is appended to the list of object numbers to be tracked Incase the mouse click does not correspond to any object a message is displayed on the screen and nothing gets added to the list of objects that are being tracked So instead of manually entering the object numbers we clicked on the objects that have to be selectively tracked AO Applications Places System GER G Y Y El Q Wed Mar 17 10 23 PM E O Visit 1120 i tee File Controls Options Windows Help 2 Window 1 Je lb aa EE EE CEA dio MPA BRAK tile FO ill 1 Click here for Zone Pick Mode DB vorts1 poly Cycle 1 Time vorts5 poly cycle 0005 varteR nahi nunla ANDA d ReOpen Replace Overlay en al AIR gt Active window Maintain limits _ Replace plots 4005 1 view data Auto update 8 2 E Active plots Hide Show Delete Draw C i Hide Show Delete Draw Select this object E Y elect this object Plots Operators PlotAtts OpAtts Variables E CH Apply Hoperators selection to all plots od j Sie Y Ba H NI LL 1 20 40 BOX Age 10 10 x EE PIER Adobe Adobe Rea I Microsoft Word ia admin miracle EI Visit 1 12 0 J EI Window 1 g 1145 pdf 7 I O Polydata plot attribu 7 TrackPoly operator a SIE 1 Polydata TrackPoly var Figure 28 Selec
47. d Feature Tracking to study the feature evolution in the dataset we are producing the visualization results for first 8 frames of this dataset 65 32 f 30 E 28 f s G 24 ViAxis 22 E f R G P Axis td 16 18 20 2 ghe 26 28 30 32 x BE Avis A nap poy B volmap poly D8 vormapt poy 16 18 20 22 24 26 28 30 32 4 d i dear DN f A Y 15 20 5 de eduunegkeend ege asar Lua wu esas rox zf 4 6 18 20 22y 2426 28 30 32 x D Figure 21 Images from Volmap Dataset 4 Indicator dataset 2D This dataset of 50 frames is 2D Dataset and each frame has a resolution of size 384 256 Each point in this 384 256 frame represents a scalar value float type Our Feature Tracking software works only for 3D Datsaset so we had to fake this dataset as a 3D dataset by adding an extra layer of Z component values The format dataset could not be read by VisIt so we had to convert it to vtk format and at the same time we added an additional Z layer to fake it is 3D The program for this is given in Appendix II Visualization results of two consecutive frames of this dataset are given below in Figure 27 Each frame now acts a 3D volume and all actions like rotating the objects selecting objects and picking objects by mouse are valid 66 DB indicator Gol Cvcle 1 Time 50 dE gt P Ko d _ 7 o KZ oot d da 300 350 DB indicator2 pol Cycle 2 Time Figure 22
48. ded li it 47 5 4 Feature Tracking amp Visualization Workflow in ViSIt sssssssussususnnnunnununnununnnnunnnnunnununnununnnnunnunnnnnnnnna 48 5 5 Modification to Feature Tracking amp Visualization Plugins ss sssssssusussununnununnununnununnununnnnunnnnnnnnnnnna 51 5 5 1 Traktable based coloring AAN dopo aaa 52 5 5 2 Selective Feature Trackingy asincisiswiued eae a Aa aiid 56 5 3 3 Picking Objects by Mouse Click id 58 5 6 Technical Challenges with VisSIt ccccsssssssssssesessesesseeseesessesaeeeseeeeesaeaeeesesaeaeeeeesesaeaseenenaeaeeenenaeaeeeeeeats 59 Gc RESUS A ttc A 62 Ta Conclusio Mereaira eaea aaasta aae aeaa a aereas ar aaae aa aaan ranap maa Aa Eadan ceauaycak doi Chaeaendunvaaaoaanivavdeubuseaees 67 A TE 68 Ro nananana saananira anaana aiaa aa aaiae aa aaa aiaa aaiae 70 Installation of Feature Tracking Visualization plugins in VisIt sssssssssussununnunsnnunnnnununnunnnnununnnnnnnnn 70 Automatic Installation 70 Manual Install accent a ee 70 Feature Tracking Visualization User manual nnniiicion ninia 71 Feature Tracking amp Extraction vii citys a i i A a N 71 MEE RN 72 Selective Feature Tracia ii 75 Picking Objects by Mouse clicks iia 76 File Fo MaS adn snsies anaes a a aa E aE aaa AAAA a e aE EAE Aaaa ustauiusiassaildnisestontuaienss 79 A EE tly 79 MA tia 80 UO CN A EA 82 tra O TOO 83 aa o e ee edo de o Ee den o 83 ee EE ee eat aa ec een ltd ee 85 Gurrenthile Aid 85 CUTpPOLlY tI teis einai ie i
49. detection of super structures and so on this framework provides a solid foundation to build upon This framework can be very useful for developing new plugins and all of Vizlab s work can be brought on a common platform which can be shared by many people across different fields 68 References 1 DA Forsyth and J Ponce Computer Vision A Modern Approach Prentice Hall 2002 2 In So Kweon and Takeo Kanade Extracting topographic terrain features from elevation maps CVGIP Image Underst 59 2 171 182 1994 3 Deborah Silver and Xin Wang Tracking and visualizing turbulent 3d features IEEE Transactions on Visualization and Computer Graphics 3 2 129 141 1997 4 Deborah Silver and Xin Wang Tracking scalar features in unstructured datasets In VIS 98 Proceedings of the conference on Visualization Op pages 79 86 Los Alamitos CA USA 1998 IEEE Computer Society Press 5 J Chen D Silver and L Jiang The feature tree Visualizing feature tracking in distributed AMR datasets In PVG 03 Proceedings of the 2003 IEEE Symposium on Parallel and Large Data Visualization and Graphics page 14 Washington DC USA 2003 IEEE Computer Society 6 Frits H Post Frank J Post Theo Van Walsum and Deborah Silver Iconic techniques for feature visualization In VIS ob Proceedings of the 6th conference on Visualization Ob page 288 Washington DC USA 1995 IEEE Computer Society 7 Deborah Silver Object oriented visualizat
50. e ah a ea hee ee ae 86 Opacity KT EE 86 APPO 1 E 88 Data structures in Scientific Visualization ccssssssssessssssssssssessseseeeesseseeseneeseseeseeseneeenenauenenenenenseeneneneeenees 88 Rekter Mesh A A Ata 88 Rectilineat M sh ul 88 Irregular Mesh Andersen tee iaiaiianchin aie wie cannes A a aca aaa eee 89 STPUCEUTEOM eS Ite satis cia cscte at ta an aaNet ead cane aries 89 Unstructured Mesa 90 Pick MOdeES a ia deele 91 AVS to vVEk CONVerteT unnnnninnnnnanaaana naaa nana nana naaa naaa aaa aaa 92 2D t 3D CONVE aaa 94 1 An Overview of the thesis The process of converting raw data into a form that is viewable and understandable to human beings is called Visualization Visualization allows us to get a better cognitive understanding of the data Scientific visualization usually deals with scientific data that has a well defined representation in 2D or 3D or has a natural geometric structure e g MRI data or wind flows Analysis of such scientific data usually provides a more intuitive interpretation for the process of hypothesis building and modeling Examples of scientific visualizations are the visualizations of intense quantities of laboratory or simulation data or the results from sensors out in the field The output of these simulations experimental and sensor data are in the form of 3D scalar datasets Figure 1 is one such example where we visualize a 2D array of numbers It very hard to interpret the meaning of these numbers on the
51. e floating values between 0 0 and 1 0 Number of entries in this file corresponds to the number of objects in the frame It looks like this colormap lt Frame number gt txt lt R1 gt lt G1 gt lt Bi gt lt alphai gt object 1 lt R2 gt lt G2 gt lt B2 gt lt alphaz gt object 2 lt Rn gt lt Gn gt lt B gt lt alphan gt object n Sample colormap file for frame 5 with 5 objects It s named as colormap5 txt 10 200 50 1 0 78 110 240 08 0 0 255 10 255 255 50 1 0 75 75 75 0 5 CurrentFile txt This file is updated every time a new file is opened by VisIt with the help of suitable database reader This is not dependent on any of the Feature Tracking plugins but is applicable to any plot or operator This file keeps a track of the information about the current file VisIt automatically creates a database for related files and this file has all the information about this database like starting and ending frame or file cycle number or the current frame or file opened number of steps to reach from start cycle 86 to end cycle If some of the files were missing in between then number of steps would not be difference of start and end cycles This file also stores the name of the database from the file name Sample CurrentFile txt CURRENT_FILE home admin data vorts1 bov PATH home admin data START_CYCLE 1 END_CYCLE 10 CYCLE 1 STEP 10 DATASET_NAME vorts curpoly txt First time this file is generated
52. e information about the dataflow network is given in Chapter 4 2 6 Examples of Scientific Visualization Visualization tools when applied to scientific data produce beautiful pictures Many scientific fields can benefit from these visualization tools We discuss application of visualization tools is some scientific fields Natural science Visualization of these phenomenon can useful for studying star formations understanding gravity waves visualizing massive supernova explosions and for molecular rendering Figure 4 shows some of these results a b 9 d Figure 4 Examples of Visualization in study of natural sciences a Star formation b Gravity plot c Visualization of massive supernova explosion d Molecular rendering Images Source http en wikipedia org wiki Scientific_visualization Geography and Ecology In this field of science visualization tools are useful for climate visualization terrain rendering and studying atmospheric anomalies in areas like Times Square Figure 5 shows the visualizations of the concepts from field of Geography and Ecology a b Figure 5 Visualization application in geography and ecology a Visualization of terrain b Climate Visualization c results from simulation framework of atmospheric anamoly around Times Square Images source http en wikipedia org wiki Scientific_visualization 18 Formal sciences In formal sciences visualization tools can benefits u
53. e surface is rendered on the screen as a solid mesh However interesting features cannot be determined by simply looking at the surface There are many variables associated with the mesh scalars and vectors Mesh being a solid hides a lot of information inside it We can discover more information about the data by applying Filters Filters are functional units that process the data to generate extract or derive features from the data Filters are attached to readers sources or other filters to modify its data in some way These filter connections form a visualization pipeline These filters can be selected by choosing a corresponding icon on the filter toolbar ParaView automatically creates this pipeline users need not worry about connecting the individual modules as in AVS Once user has finished selecting the appropriate filters the results are rendered on the screen on clicking Apply Button ParaView does not give internal details about the visualization pipeline There is no information on how the pipeline is formed methods that would be executed or how the modules are connected ParaView also lacks proper documentation for process of adding new filters or readers 2 5 3 Visit The VisIt 11 project originated at Lawrence Livermore National Laboratory as part of the Advanced Simulation and Computing ASC program of the Department of Energy s DOE National Nuclear Security Agency but it has gone on to become a distributed project be
54. eFile void In case of database reader the following method needs to be implemented vtkDataSet avtPolyFileFormat GetMesh const char This method is described in avt lt Databasename gt FileFormat class and changes in here would accomplish the task 5 Then Compile and run 40 5 Feature Tracking amp Visualization in Visit 5 1 Motivation Vizlab has pioneered the use of Feature Tracking for analyzing time varying datasets Feature Tracking was incorporated as module in AVS Since AVS is a proprietary software the license costs limits its widespread usage Also renewal of AVS license was very costly Thus new avenues were sort for and Open Source Visualization Software Packages were sought to replace AVS Among the Open Source Visualization Software Packages VisIt was chosen as the best option to port Feature Tracking VisIt s rich feature set its support of distributed and parallel processing and its architecture made it a good choice Extending new capabilities like the Feature Tracking Visualization in VisIt would add a lot more value to VisIt and benefit the users tremendously Previous work 23 in porting Feature Tracking into VisIt was successful in extracting and tracking features from a time varying dataset But the visualization plugins were found to be incomplete and incompatible with newer version of VisIt Hence the aim of this thesis is to e Successfully implement Feature Tracking and Visualization cap
55. ed Mar 17 10 24 pm U Save Picks as Variables default Float Format g 3 Concise Output P Show Mesh Name P Display incident nodes zones 3 Display global nodes zones P Display reference pick letter Display for Nodes P Show Timestep E ld _ Physical Coords 3 Domain Logical Coords Block Logical Coords Display for Zones E ld 3 Domain Logical Coords Block Logical Coords P Automatically show window 3 Don t clear this window Clear Picks NN Opacity For Other Objects SE 2 CAxis Ra N 3 Create time curve with next pick Repeat Pick Make default Reset LL LL DS ESE es 20 40 60x Ai 100 12 P Apply Post Dismiss x Create spreadsheet with next pick Display in Spreadsheet Make default user d E 8 Adobe Adobe R Microsoft word E adminemiracie vist 1 12 0 window 1 tg 1145pdf ID Polydata plot attr TrackPoly operat Pick Figure 29 Action on selecting an object by Mouse Click A new window titled Pick pops up This is PickAttribute window which gives a detailed information about the selected object and lets the user chose any specific action from the attribute window The object is automatically labeled A Another change which takes place is addition of the object number to the textbox ObjectNumberToTrack in PolyData plot attribute window 79 File Formats poly This file is written
56. ethod is a Frame type dataset which contains the information of number of objects in a frame and mass volume of each object in the frame 5 5 1 2 Modifications to Poly Database Reader In Poly Database reader we made changes to the following method vtkDataSet avtPolyFileFormat GetMesh const char This method has the name of the poly file that is passed as an argument So in the method we can get file name and frame number The task of this method is to read poly file and generates a mesh from those data values In this process it finds out the number of the objects in the frame We add code to put the file name and number of objects for that file or frame in textfile called Curpoly txt Earlier curpoly txt was having just name and number of objects In this thesis we add an extra line to represent the frame number that initially starts from 0 and is modified by the TrackPoly operator for each frame 5 5 1 3 Modifications to TrackPoly Operator The method modified in TrackPoly operator is from avtTrackPolyFilter class and defined avtTrackPolyFilter C class The signature of the modified method is vtkDataSet avtTrackPolyFilter ExecuteData vtkDataset int std string 55 This is auto generated method had to be reimplemented for the operator to function properly In this method we perform the following actions e Get the current path from operator attributes e Open the curpoly txt file this file would be at the cur
57. f feature tracking with proper color codes The figure shows 3 consecutive frames of vorts data Lets study the evolution of object A and Object B in a b and c The necessity for assigned coloring based on track tables becomes evident in c as its clear the object A has split into two and both the parts retaining the same color 44 5 3 Custom Plugins The Feature Tracking amp Visualization project in VisIt has 4 plugins These plugins are categorized into two groups The first group is responsible for Feature Extraction and tracking is named as Feature Track operator The second group has three plugin Poly TrackPoly PolyData for visualizing the results The descriptions of these plugins are as follows 5 3 1 Feature Extraction amp Tracking Group 5 3 1 1 FeatureTrack operator This operator extracts meaningful features from raw data and tracks them In 23 this module generated poly attr uocd trak and TrakTable files In addition to those files we generate one more file per frame in this module This file has colormap information for objects in the frame and the format of the file can be found in Appendix I Figure 15 Feature Tracking Plugin development in Visit 5 3 2 Visualization Group 5 3 2 1 Poly Database reader This plugin reads the information in poly files gathers metadata about the dataset and creates a vtkDataset object to be passed on to the network This plugin generates a file curpoly
58. fluid flow biological etc where emphasis is on realistic rendering of volumes surfaces illumination sources and so forth perhaps with a dynamic component of time Scientific Visualization is the use of data driven computer graphics to aid in the understanding of scientific data Is scientific visualization just computer graphics then Computer graphics is the medium in which modern visualization is practiced however visualization including scientific visualization is more than simply computer graphics It uses graphics as a tool to represent data and concepts from computer simulations and collected data Visualization is actually the process of selecting and combining representations and packaging this information in understandable presentations 2 1 Underlying principles In this section we will discuss various reasons for using scientific visualization and the effect they have on scientific experiments Also we will have a look at some basic steps of visualization Scientific visualization is form a communication and we need to be very clear of our audience as they have to grasp what happens to the information as it passes from numbers to pictures 2 2 The Goal of scientific visualization Scientific visualization is employed as a tool to gain insights into natural processes easily For instance the goal might be to demonstrate a scientific concept in which case the presentation to scientist would be different from a presentation
59. frame tracking information 84 gets appended at the end of the file trakTable gives the evolution history of each object frame by frame along with an indication of any events that occur The format of the file is like this lt Frame gt lt Objects of Currents Frame gt 1 lt Corresponding objects in next frame gt T acts as delimiter to help identify between objects of current frame and next frame If the objects of the current frame are disappearing in next frame there is nothing to the right of 1 If new objects are forming in next frame then there would nothing to left of 1 Sample trakTable file Frame 3 1 5 7 1 2 5 3 object from frame 2 1 5 7 combine as objects 2 5 in frame 3 6 1 1 10 Object 6 from frame 2 splits into Object 1 10 in Frame 3 2 1 3 object 2 continues as object 3 in next frame 4 1 4 1 6 1 7 this is a new object created in Frame 3 1 8 1 9 3 1 this object no longer continues in Frame 3 it dissipates 85 colormap txt This file is generated for each frame of the dataset VisIt allows to assign different colors to objects in the frame than the ones defined in poly files This information has to be provided to the Visit and it constructs a color map out of this information So this file contains red R green G blue B intensities and alpha values opacity for each object R G and B are integer values in the range 0 to 255 and alpha values ar
60. g objects with mouse proved to be very useful to remove the visual clutter and focus on features of interest The results from these datasets are given below 1 vorts dataset This simulation of 100 frames or timesteps was generated on a Connection Machine CM5 for studying problems of Computational Fluid Dynamics CFD 3 The dataset contains vorticity and scalar vorticity magnitude is computed for visualization and tracking This simulation data is courtesy of N Zabusky and V Fernandez 27 28 Resolution of each frame is 128 128 128 and at each point a scalar value float type is stored The results for application of feature tracking to visualize frames 5 to 10 of this vorticity dataset are given below in Figure 24 The visual analysis is possible because of a consistent coloring across frames assisting the user in tracking objects It s very clear from the results that the one of the objects continues encircled and labeled as continuation in figure 24 in all frames Also two objects from frame 7 combine to form a bigger object a new object is formed in this frame which continues to grow in the rest for the frames 63 New Object formed 0 2 by xA 00 0 2 4 OAA 00 a Continuation Dem x 0 AT me x Dissipation Figure 19 Visualization Results for Feature Tracking on vorticity dataset vorts from frames 5 to 10 We can clearly see the object encircled in frame 5 continuing in the rest of
61. ground YSAxis Y Axis V lt Axis 0x Axie 100 8 20 4 SON Ab 200 120 Gi 20 40 SI Ai a b c YsAxis YiA V scle L Axis EN ANA la Ss 90 xa ke L ad 20 40 50X A 8 X A A wo d e DI Figure 27 Selective Feature Tracking and Visualization for an Object Number 5 from vorticity dataset vorts frames or timesteps 1 to 6 The object is selected via the GUI described in Figure 1 The opacity for the object number 5 is 100 and rest for the rest of the objects opacity is set at 5 Picking Objects by Mouse clicks Selective feature tracking is very useful feature but it becomes particularly difficult to know the object numbers of each and every object in a frame It s easy to get object numbers by using Label plot and then manually entering the object numbers in the text box of the objects to be tracked But it s not 77 practical and takes a lot of time to figure out the object numbers The most intuitive behavior in such situations is is to be able to select objects by Mouse Clicks This functionality is added to the Feature Tracking module It s very simple and easy to select objects by Mouse clicks see Figure 22 First a pick mode has to be chosen This can be either Zone Pick or Node Pick Mode and whenever an object is selected a new Pick window pops up see appendix II for more info on pick modes Another change that can be seen is in PolyData plot attribute window The textbox asso
62. guage Extensible with dynamically loaded plug ins VisIt achieves extensibility through the use of dynamically loaded plugins All of VisIt s plots operators and database readers are implemented as plugins and are loaded at run time from the plugin directory From Data to pictures in VisIt The process of conversion of data to pictures in Vislt is very similar to the process in other visualization software packages The difference comes in the underlying network that is created VisIt automatically creates an AVT network for the user depending on the action performed by the user If the user performs the following actions 1 Load a data file 2 Apply Operator or filter to the data if any And user chooses splice operator 3 Choose a plot contour plot 4 Execute the network and draw the results on the screen Rendering filters added automatically by avtPlot Figure 3 VisIt Dataflow network Image source http visitusers org index php title AVT_Overview 17 Then a network shown in Figure 3 is generated automatically showing contribution of each step in the entire network VisIt keeps on adding to this network according to the user action and only when user chooses Apply the network is executed and result rendered on to the viewer screen VisIt provides plugins corresponding to each of the action There are many built in plugins for handling database i e loading files operators and plots Mor
63. ing developed by several groups VisIt is an open source turnkey application for large scale simulated and experimental data sets Its charter goes beyond pretty pictures the application is an infrastructure for parallelized general post processing of extremely massive data sets Target use cases include data exploration comparative analysis visual debugging quantitative analysis and presentation graphics VisIt leverages several third party libraries like the Qt widget library 12 the Python programming language and the 15 Visualization ToolKit VTK library 13 for its data model and many of its visualization algorithms VisIt has been ported to Windows Mac and many UNIX variants including AIX IRIX Solaris Tru64 and of course Linux including ports for SGI s Altix Cray s XT4 and many commodity clusters Some of the key features of VisIt are listed below e Rich set of features for scalar vector and tensor visualization Vislt s visualization options can be broadly classified in two main categories as mentioned in the VisIt Developer Manual 20 1 Plots to visualize data and include boundary contour curve mesh streamline subset surface tensor vector 2 Operators consists of operations that can be performed on the data prior to visualization like slice index onion peel iso surface etc e Qualitative and Quantitative visualization VisIt is also a powerful analysis tool It provides support for derived
64. ion IEEE Computer Graphics Applications 15 3 54 62 1995 8 Michael Friendly 2008 Milestones in the history of thematic cartography statistical graphics and data visualization 9 Keates J F Understanding Maps 10 Scientific Visualization http wikipedia iki 11 VisIt Homepage https wci lInl gov codes visit 12 Qt Widget Library http qwt sourceforge net 13 Visualization Toolkit Homepage http www vtk org 14 Wikipedia description of Client Server model http en wikipedia org wiki Client server 15 An overview of Remote Procedure Calls RPC http www cs cf ac uk Dave C node33 html 16 What is a Proxy Class in C http stackoverflow com questions 994488 what is proxy class in c 17 L Jiang Output sensitive feature tracking Master s thesis Rutgers The State University of New Jersey 2005 18 H Freeman and S P Morse On searching a contour map for a given terrain profile Journal of the Franklin Institute 1967 p 248 1 25 19 High level design of Vislt http 20 VisIt Use Manual https wcillnl SES SEA Aleida 5 pdf 21 AVS Express Homepage http www avs com software soft t avsxps html 22 Vislt supported File Formats wikipedia page https visualization hpc mil wiki VisIt Data Formats 23 R Pangrikar FEATURE TRACKING OF 3D SCALAR DATASETS IN THE VISIT ENVIRONMENT Master s thesis The State University of New Jersey 2008 24 Map data structure description http
65. ion Software Packages cnn 10 A En 10 REENEN 12 DOS VI EE 14 2 6 Examples of Scientific Visualization coccion rr 17 3 Fe t re Tracki g racer rd 19 3 1 Feature Tracking Techniques omnia een errores 20 3 2 Feature Tracking Algorithms at Vizlab ocnnniicinn nn 21 3 3 Applications of Vizlab s Feature Tracking Algorithms acciones 25 3 4 Software Implementations of Feature Tracking Algorithms omnes 26 3 4 1 AVS Express implementation ENEE 26 3 4 2 Distributed Feature Tracking Implementation EEN 27 3 4 3 Vislt Implementationianiais aati aan a aia ace Sie dane ae tie 28 A How Vist Wo ES ui die 29 4 1 High level design Of Vislt c c cscssesecccenceseseeessecseeeceestecesenesnenecenesenseceeenesseceeeseerseeeeneneneneeneneneeaneenenensesees 29 4 2 Connectivity Communication between COMPONE DUS mmm 30 4 3 Worktlow OF VIS E aa a aaa aaa ara a ii aaia 31 noz 35 4 5 Adding new plugins in ViSIt ncincninncininmnnin anunn nnn n 36 5 Feature Tracking amp Visualization in ViSIt onninniniinni res 40 Nee VC lu EE A0 5 2 Feature Tracking amp Visualization FunctionalitiesS s sssssssusussununnnnunnunnnnnnunnunnnnnnunnununnnnunnnnunnnnnnnnnnnna 41 5 2 1 TrakTable based Color coding iii io 41 5 2 2 Selective Feature TCM a ria 42 5 2 3 Picking Objects by Mouse Click ee 42 5 3 CUOStOM Plus 0 aii 44 5 3 1 Feature Extraction amp Tracking Group miii ee 44 AER tee E 44 5 3 3 A t Generat ed Fleege diia 45 5 3 Newly ad
66. ir own but when presented as a picture we get a better idea of what the data is representing 0265640 132304 133732 032051 037334 024721 015013 052226 001662 Density 0265660 025537 064663 054606 043244 074076 124153 135216 126614 am om ia zem 0265700 144210 056426 044700 042650 165230 137037 003655 006254 i 0265720 134453 124327 176005 027034 107614 170774 073702 067274 0265740 072451 007735 147620 061064 157435 113057 155356 114603 0265760 107204 102316 171451 046040 120223 001774 030477 046673 0266000 171317 116055 155117 134444 167210 041405 147127 050505 0266020 004137 046472 124015 134360 173550 053517 044635 021135 0266040 070176 047705 113754 175477 105532 076515 177366 056333 0266060 041023 074017 127113 003214 037026 037640 066171 123424 0266100 067701 037406 140000 165341 072410 100032 125455 056646 0266120 006716 071402 055672 132571 105645 170073 050376 072117 0266140 024451 007424 114200 077733 024434 012546 172404 102345 0266160 040223 050170 055164 164634 047154 126525 112514 032315 0266200 016041 176055 042766 025015 176314 017234 110060 014515 0266220 117156 030746 154234 125001 151144 163706 136237 164376 0266240 137055 062276 161755 115466 005322 132567 073216 002655 0266260 171466 126161 117155 065763 016177 014460 112765 055527 0266300 003767 175367 104754 036436 172172 150750 043643 145410 0266320 072074 000007 040627 070652 173011 002151 125132 140214 0266340 060115 014356 015164 067027 120206 070242 033065 131334 0266
67. is method calls NetworkManager AddFilter which does the following actions 1 The proper plugin type is loaded 2 An avtFilter is instantiated and registered with a workingNet 3 The attributes of the filter are set e Setting up the plots This is similar to the setting up of operators The following actions take place 1 RPCExecutor lt PreparePlotRPC gt Execute is called This must be called first to instantiate the correct type of attributes so that the subsequent call to MakePlot will be able to load the attribute values into this instance 2 RPCExecutor lt MakePlotRPC gt Execute is called 1 NetworkManager MakePlot is called 1 The proper plugin type is loaded 2 An avtPlot is instantiated and registered with a workingNet 3 The attributes of the plot are set 2 An Id is obtained from the network manager and returned to the viewer This Id is used to refer to this plot in the future For picks etc e Executing the network RPCExecute lt ReadRPC gt Execute is called This methods calls the following two methods 35 1 NetworkManager GetOutput is called 1 Each module of the pipeline is connected 2 DataNetwork GetWriter is called avtPlot Execute is called The return may be either geometry or a NULL object saying that we need to kick into Scalable Rendering mode 2 The output is sent through a socket with a WriteData call The output comes as an avtDataObjectWriter This output may be either a data
68. ithm 37 for communication between processors Once the distributed features are extracted on different processors tracking server operates sequentially to get tracking results A partial merge strategy was also proposed where processors communicate with their neighbors to determine the local connectivity 3 2 3 Adaptive Mesh Refinment AMR Feature Tracking Algorithm Chen et al 5 describe a distributed feature extraction and tracking process for AMR datasets In AMR datasets we have grid points with varying resolutions and features can span across multiple grid level and processors So the tracking must be performed across time across levels and across processors Tracking is computed temporally across lowest grid level and then computed across spatial levels of refinement Features 25 formed in higher level are tracked in subsequent time step as Feature Trees Please refer to 5 for more information on AMR Feature Tracking 3 3 Applications of Vizlab s Feature Tracking Algorithms Feature tracking can be applied to any time varying 3D dataset At Vizlab Feature Tracking was applied to many real life engineering applications The structure of the data in these datasets was varied Structured Mesh Unstructured Mesh etc as in appendix II Some of the broad application areas are as follows 1 2 3 4 Meteorology Feature tracking was applied to the cloud water simulation This simulation consisted of
69. k in hand This is sometimes still done with large time varying datasets but now mostly people use off the shelf softwares with some modifications to perform a particular task 2 5 Scientific Visualization Software Packages Visualization Software package sometimes referred to as dataflow package is the mostly prominently used visualization tools These are modular softwares are based on Object Oriented Programming languages that facilitate addition of new capabilities as modules In this section we discuss about 3 different software packages that are widely used Each of these packages creates a network that is executed to produce visualizations 2 5 1 AVS Express AVS Express is a comprehensive and versatile data visualization tool for both non programmers and experienced developers It provides powerful visualization methods for challenging problems in a vast range of fields including science business engineering medicine telecommunications and environmental research AVS Express enables object oriented development of rich and highly interactive scientific and technical data visualizations for a wide range of computing platforms www avs com AVS Express has the following attractive features e Object Oriented AVS Express development approach is object oriented it supports the encapsulation of data and methods class inheritance templates and instances object hierarchies and polymorphism In AVS Express all application compo
70. lying file formats VisIt has a database reader for each supported file format and the database reader is a plugin that reads the data from the input file and imports it into VisIt 22 If VisIt does not support your data format then you can first translate your data into a format that VisIt can read e g Silo VTK etc or you can create a new database reader plugin for Visit 4 5 Adding new plugins in Vislt VisIt comes with a graphical plugin creation tool which greatly simplifies the process of creating new plugins The user describes the properties of the plugin and then the tool generates most of the code necessary to implement the plugin The only code you need to write is the C code that actually performs the operation Steps for creating a new plugin VisIt provides with some tools to help developers in adding any new functionality These tools can be found under the following location lt visithomefolder gt src bin gt and more information about these tools can be found in VisIt s User Manual 20 If this folder has been added to the system path then the tools in this folder can be accessed from any location by just typing the name of the tool Otherwise full path has to be mentioned For example we want use xml2edit tool in some folder say home admin Then the commands to run this tool are cd home admin xmledit lt visithomefolder src bin gt xmledit 37 For the rest of commands in this thesis we assume that
71. meta data about a file allowing the user to set up plots and operators without an engine Engine The engine performs data processing in response to requests from the viewer There are both parallel and serial forms of the engine called engine_ser and engine_par respectively The engine sometimes performs rendering although it is also performed on the viewer Table 1 Visits multiple separate processes Source This information is taken from http visitusers org index php title High_level_design 4 2 Connectivity amp Communication between components The connections between the various components are shown in the figure 12 At the lowest level the communication is done with sockets However two separate layers are built on top of that e The first is for exporting state The viewer keeps all of its state in various instances of Vislts AttributeSubject class UI modules such as the GUI and CLI subscribe to this state refer to 25 by Gamma et al for more details Thus when state changes on the viewer the AttributeSubjects automatically push this state out to its subscribers The second is for remote procedure calls RPCs 15 When a component wants another component to perform an action it issues an RPC o The RPCs come via a proxy class 16 For example there is a module named ViewerProxy Both the GUI and CLI link in ViewerProxy and make method calls to this class Each method call becomes an RPC
72. nal space without specifying the relations between those dimensions The visualization system learns to extract and track features in complex 4D flow field according to their visual properties like the location shape and size 21 3 2 Feature Tracking Algorithms at Vizlab At Vizlab the Feature Tracking algorithms for 3D Scalar Fields are based on a framework as shown in Figure 12 The goal of the process is to obtain dramatic data reduction and thus help scientist quickly focus on a few features or events of interest The major steps are e Feature Extraction The first step is to identify and segment features of interest from the dataset to be tracked The method used depends on the definition of a feature which can differ from domain to domain Usually features are defined as threshold connected components 6 7 3 Raw Datasets SE Ep 1 i Enhanced Feature Extraction il LA 1 Visualization Feature Tracking A Y WM Th H Event Querying Feature Isolation y JSZ Figure 9 The Feature Track based visualization pipeline Image source 6 Feature Quantification e Feature Tracking In this step the evolution of the extracted features is followed over time noting various events that occur during their evolution e Feature Quantification Once features are extracted they are quantified and information about them e g mass centroid etc can be calculated Enhanced visualization and e
73. nents from the lowest to the highest level are objects 11 e Visual development The Network Editor is AVS Express main interface It is a visual development environment that is used to connect define assemble and manipulate objects through mouse driven operations e Visualization application AVS Express provides hundreds of predefined application components objects that process display and manipulate data The objects and application components that you connect and assemble in the Network Editor control how data is processed and how it is displayed Furthermore AVS Express also provides programming interface APIs to C C and Fortran allowing developers to easily integrate their own modules into AVS Express From Data to Pictures in AVS To transform data to Pictures in AVS one must follow these steps 1 Import the data in AVS 2 Process the Data if needed 3 Apply one or more Visualization techniques 4 View the results AVS has many built in module for performing all of the above mentioned tasks For e g ReadField module importing the data from fld file into AVS Downsize modules does the processing of the data and as the name suggests it does some processing based on some criteria provided by the user Similarly there are modules to apply visualization techniques and view the results A user selects the appropriate modules manually and builds a network as shown in Figure 6 AVS facilitates the process of ne
74. of the important features of ParaView are given below e Visualization Capabilities ParaView handles structured unstructured polygonal multiblock and AMR data types All the processing or filtering operations produce datasets ParaView can 13 be used to inspect vector field by applying glyphs extract contours and iso surfaces cut or clip regions by clipping planes or generate streamlines using constant step or adaptive iterators The points in a dataset can be warped with scalar or vector quantities Python programming interface can be used for advanced data processing Input Output and file formats ParaView supports a variety of file formats ParaView reader and writer provides a complete list of supported file formats User interaction Qt application framework introduces flexibility and interactivity Parameters on the filters can be changed by directly interacting with the 3D view using 3D manipulators Interactive frame rates in maintained by using LOD level of detail models Large data and distributed computing ParaView runs parallelly on distributed and shared memory systems using MPI These include workstation clusters visualization systems large servers supercomputers etc ParaView uses the data parallel model in which the data is broken into pieces to be processed by different processes Most of the visualization algorithms function without any change when running in parallel ParaView also supports ghost levels used
75. olormap file for that frame If current frame number is 2 then create and open a file called colormap2 txt 4 Allocate a vector lt string gt to store the color information for Frame1 previous frame and Frame2 current frame objects 5 Open colormap file for previous frame and fill in Frame1 vector from those values The number of lines in colormap indicates number of objects in previous frame and that is the size of Framel vector Each entry in the vector have R G B and alpha values in it as one string separated by blank spaces 6 Open the trakTable file and fill TrakTable vector from the entries corresponding to the current frame Each element of TrakTable vector is a string Use ReadCurrentFrameTrackInfo routine for this purpose 7 Tterate for the number of entries in TrakTable vector and for each iteration e Parse each element of TrakTable vector which is a string For example 3 1 2 can be one of the entries 1 is the delimiter whatever is to the left of 1 is previous frame object and the number on the right represent current frame objects e Separate out previous frame and current frame objects e Set the color for current frame objects of Frame based on previous frame colors from Framel vector i e for3 1 2 object2 of current frame Frame2 should get the color object3 from previous frame Frame1 e Ifthere are more than current frame objects set the same color for all those objects e If more than one objects from
76. ome interesting phenomenon on the screen and decides to see feature from Frame 25 onwards Scientist wants the visualization starting from a different frame In AVS there is no separation between extraction and visualization stage then entire process has to be repeated Even though there isn t any change in data just because visualization part cannot be viewed in isolation the extraction and segmentation part is repeated For smaller datasets this might not be a point of concern But for large datasets a lot of time and resources are wasted on repeating the process on data that did not change But Visit is different since it separates Feature Extraction Tracking from Visualization there is no repetition of work The scientist performs feature extraction and tracking only once Visualization can now be repeated as many times as the scientist wishes to As features extraction is not repeated again and again visualization becomes much more efficient on larger datasets 5 2 Feature Tracking amp Visualization Functionalities 5 2 1 TrakTable based Color coding From figure 14 it is clear that objects in each frame on the left side of the figure have colors assigned to them arbitrarily There is no correlation between the objects and colors assigned to them In time varying datasets an object from one frame can move around in the next few frames of the dataset It can change its shape and size splits into smaller objects form a bigger object
77. ood practice for scientists to question the accuracy and the validity of all the information that is presented to them It s always important to get this feedback and make changes to the process in order to get proper and accurate results in visualization for the intended audience 2 4 Scientific Visualization tools A number of tools are available for creating visualization of information These tools can be categorized as Plotting libraries Software libraries were developed that enabled researchers to generate charts graphs and plots without the need for reinventing the graphics themselves Since the form of interaction is through programming it has limited interactivity Turn key packages A turnkey visualization package is a program designed specifically for doing visualization and contains controls widgets for most options users would want to exercise when visualizing data This is accomplished through the use of pull down menus or popup windows with control panels Examples are Vis 5D Gnuplot etc 10 e Dataflow packages These are designed as tools to be used directly by the scientist The dataflow concept consists of breaking down the tasks into small programs each of which does one thing Each task is represented as a module Examples of this kind of packages are softwares like AVS e Writing your own softwares before dataflow packages and other tools were available the programs were customized for a particular tas
78. operator attributes individually one by one Select the corresponding plot and operator and fill in the attributes 5 Click on Draw to get the following visualization of vorts1 poly file 75 Selective Feature Tracking Figure 20 shows PolyData plot and TrackPoly operator attribute windows These attributes are applicable to all frames or timesteps of vorts dataset When a user wants to track only some of the objects then the rest of the objects fade away The steps involved for this process are 1 Click on Yes checkbox for the label SelectiveObjTracking 2 Then enter the object numbers for the objects which you want to track This can be entered either manually if you know the object number or you can double click on an object and its object number would be extracted automatically Note For this happens because we added that functionality while designing the plugin and VisIt should be either in zone pick or node pick mode next sextion has more details about it 4G Applications Places System CR E W E l Mon mar 8 6 18 Pm 0 D Visit 1 12 0 Window 1 File Contols Options Windows EFE EE CERA UR K fal Polydata plot attributes P Legend E Lighting 3 Use min PolyFlePath GENERATED_TRACK_FILES yee max _InitialcolorScheme Random v UserSelectedFile Scale EASA SelectedFile Min Active window Maintain limits _4 Replace plots 1 iview data Autoupdate Make defaul
79. other component and its best if VisIt to figures the right environment and generates the appropriate makefile 5 3 4 Newly added files In addition to modifying auto generated files some new files were also added Table 3 gives a list of all the files that have been added in Feature Tracking plugins Some of the files are marked with an it means that these are auto generated files and they have been modified to add new methods or change the functionality of the existing method The new files are added to the lt pluginname gt directory VisIt is so structured that all the plugins either existing in Vislt or new ones are grouped in lt visithomefolder src gt directory under the categories of plot operator or databases So a file in FeatureTrack operator column is added to lt visithomefolder src operators FeatureTrack gt The new files added can be either header file h or implementation files C or both In the table there are files like ObjSegment stRGB h which means the file name is stRGB h and it s under a folder called ObjSegment For FeatureTrack operator plugin two new methods have been added to avtFeatureTrackFilter class Figure 14 details of which are given later on in this section There has 48 been no change to the rest of the files mentioned in table 2 for Feature Track Operator hence they are retained as it is from previous work 23 FeatureTrack Operator Polydata Plot TrackPoly Operator Poly Datab
80. package has defined a set of data types upon which the algorithms depend to function Now we will see the basic data structures and the way visualization packages sees them Uniform Mesh A uniform mesh has grid spacings that don t change along the axes To specify the grid along one axis any three of the four values min value max value step size and number of steps must be specified and we can find the last value With three of the above values for each axis the software package can construct the mesh and generate connectivity information Uniform 2D Mesh Uniform 3D Mesh Uniform Mesh Image source http sciviz aict ualberta ca index php page data amp sub_page structure Rectilinear Mesh A rectilinear mesh has orthogonal axes and the grid coordinates are specified along each axis for 2D datasets we must provide X Y coordinates and for 3D datasets Z coordinates should also be included 89 Rectilinear 2D Mesh Rectilinear 3D Mesh Image source http sciviz aict ualberta ca index php page data amp sub_page structure Irregular Mesh An irregular mesh has less symmetry than uniform or rectilinear meshes The coordinates of all points in the mesh must be provided to the visualization software however the connectivity between the nodes can be inferred Irregular 2D Mesh Irregular 3D Mesh Image source http sciviz aict ualberta ca index php page datagsub _page structure Structured Mesh
81. ports One thing to note here is that along with the path the dataset name is also to be provided If home data vorts is entered it means under the directory home data there is vorts dataset The dataset would have files for each frame and would typically be numbered as vorts1 bov vorts2 bov and so on e initialTimestep this is the frame number or timestep from which object segmentation has to be started e finalTimestep the frame number until which objects have to be segmented and tracked e percentThreshold the threshold value for segmentation Everything above the given threshold value would be segmented and the rest would be ignored e smallestObjVolToTrack this specify how small can be the smallest object that is being tracked The value of 1 indicates that objects as small as having only one node are also displayed e timestepPrecision sometimes the datasets are generated by some simulations and its quite possible to have a lot of frames in them ranging thousands In such cases the numbering of 72 the datasets can be vorts0001 poly vorts0002 poly and so on In this precision is 4 In case of precision 1 the numbering of the dataset is normal without adding any zero s to it Featurelrack operator attributes inputBaseFilePath nome data vorts initialTimeStep finalTimeStep timestepIncrement percentThreshold smallestObj VolToTrack fi timestepPrecision fi Make default Reset Apply Post
82. previous frame combine use find_merge_color routine to get the color of the heaviest object as the final color e Ifthe current frame objects are newly created then generated another set of RG and B values randomly and set alpha as 1 8 After step7 Frame2 vector is populated with colors based on Frame colors write these colors to corresponding colormap file and close the file 2 void avtFeatureTrackFilter ReadCurrentFrameTrackInfo string int vector lt string gt amp 54 This helper method or routine would parse the TrakTrable file and extract the information for the current frame The format of trakTable file is given in Appendix I So if the current frame number is 2 this method will extract all the lines starting from Frame 2 in the file till next frame i e Frame 3 All the lines are written to a vector of string named TrakTable while skipping blank lines if any 3 vector lt string gt avtFeatureTrackFilter Tokenize const string amp This method or routine breaks a string into tokens in this case delimiter is blank space This is useful on the TrakTable vector where line has certain number and this routine breaks into separate tokens and process them 4 string avtFeatureTrackFilter find_merge_color Frame amp vector lt string gt This method or routine find the mass of the heaviest object among the given list and returns the color corresponding to the heaviest object One of the inputs of this m
83. ps monitoring these are the attributes of the plot Whenever there is a change in attributes the plot is notified of it Now to make our plot PolyData aware of the Pick Attribute changes we need to one more subject i e PickAttributes in the plot s monitoring list Whenever PickAttributes change the object number is extracted from PickAttributes by a method in PolyData plot s UpdateWindow method This extracted object number is placed in the textbox for selective feature tracking as described in previous section In previous section user entered this information by hand now the process is automated Once the list of selective feature tracking objects is formed the user clicks on Apply and from there on it follows the same procedure as in previous section to selectively visualize those objects 59 5 3 3 1 New Methods in PolyData Plot 1 void UpdateObjectPicked bool Whenever an object is picked this method updates the ObjNumToTrack widget Textbox in PolyData plot attributes with the current object number It appends the current object number to already existing list of objects to be tracked 2 int PickedObject PickAttributes has lot of information about the object that is picked This method extracts the object numbers from the whole information and returns an integer representing object number to the calling method Note Vislt starts object numbering from 0 whereas our numbering starts from 1 so to main
84. ques for extracting and tracking features from time varying datasets These techniques can be broadly classified as e Overlap based Feature Tracking Silver amp Wang 30 4 presented a feature tracking technique that extracts features organizes them into an octree structure and tracks the threshold connected components in subsequent timesteps by assuming that all the features between adjacent timestep overlap Later they improved the implementation by tracking features in unstructured datasets e Attribute based Feature Tracking Reinders et al 31 described a tracking technique that tracks features based on attributes like mass position and size A number of correspondence functions are tested for each attribute resulting in correspondence factor Correspondence factor makes it possible to match likely feature across subsequent frame in a dataset e Higher dimensional isosurfacing based Feature Tracking Ji et al 32 introduced a method to track local features from time varying data by analyzing higher dimensional geometry Rather than extracting features such as isosurfaces separately from individual frames of a dataset and computing the spatial correspondence between the features correspondence between the featues can be obtained by higher dimensional isosrufacing e Machine learning based Feature Tracking Tzen and Ma 33 present a machine learning based visualization system to extract and track features of interest in higher dimensio
85. ramework to work on This can be done as follows cd cd lt visithomefolder src lt plugintype gt lt pluginname gt gt xml2plugin lt pluginname gt xml The framework depends on the type of plugin The files generated by VisIt depend on the type of the plugin Table 1 gives a list of files generated for Feature Tracking plugins These plugins have to be compiled before using and some of the methods in these files have to be edited to compile Again the methods to be edited modified depend on type of plugin 38 XMLEdit untitded xmi Makefile Attribute Enums Fields Functions Constants Includes Code Plugin C Attribute only General Plugin attributes Plugin type M Plugin is enabled by default Name Label Version F Has icon rPlot Plugin attributes T Mesh F Scalar T Curve T Tensor Variable types accepted by the plot T Vector F Materia F Subset T Species T Symmetric Tensor T Label T Array r Database Plugin attributes Database type Extensions T File format can also write data T File format provides options for reading or writing data T Format uses explicit filenames Filenames Figure 13 Untitled xml file is generated when the user runs xmledit This file has to be saved with the same name as that of the plugin after filling in the information While adding a plot the important methods are virtual void virtual avtMapper vir
86. rent We assume that the user knows the object numbers of all objects in a frame and selects the appropriate object by writing those numbers in the list This assumption becomes unviable as the number of objects in a frame increase To assist the users in such cases we make our implementation intuitive by selecting the objects at the click of a mouse 5 2 3 Picking Objects by Mouse Click When presented with a lot of data on the screen it is intuitive for the user to click on a particular region on the screen and get more information about that region VisIt extends this concept by letting users know more information about the objects rendered on the screen When a user clicks on any object on the screen in Node Pick or Zone Pick mode a Pick window pops up that has all the additional details about the object We extend this concept to build a list of objects for selective tracking A user clicks on all the objects of interest the plugin would automatically extract the node number and build the list 43 User would still be able to edit and make changes to the list manually Once user is finished with this operation they should click Apply Now only the objects mentioned in the list are rendered and rest of them are made transparent 3120 wa Y Axis 60 340 Object B 40 b Figure 14 The images on the left half show random colors being assigned to objects in the frame on the right shows the results o
87. rent path that user enters as an attribute to the operator e Modify the last line representing the frame number by adding 1 to the existing value e Close the file Why modify curpoly txt from TrackPoly rather than Poly reader The files are automatically grouped by VisIt into a database that could be played as an animation Clicking on Open or double clicking the file loads the first file in the database From there on other files are loaded by clicking on animation control buttons forward reverse and play The curpoly txt is generated by Poly plugin The method generating curpoly txt is not called when the files were loaded from animation control panel So even though file was being generated it was not updated for every frame We noticed that the operator methods are called every time whether the file is played as part of animation or its loaded by double clicking it so we moved the task of updating the curpoly txt file to TrackPoly operator method Poly reader plugin just creates the file and TrackPoly plugin just updates it for every time frame To double check we introduced another file CurrentFile txt This file contains a lot more information and can replace some of the attributes that users need to fill in Details of this file are given in Technical challenges section of this chapter 5 5 1 4 New methods in PolyData plot For TrakTable based coloring few methods have to be added or modified in avtPolydata class
88. s an improvement over the previous work that focused on Feature Extraction in Visit In this thesis the emphasis is on the visualization of features by assigning a constant color to the objects that move or change their shape over a period of time Our algorithm gives a scientist and users an option to choose only the objects of interest amongst all the extracted objects Scientists can then focus their attention solely on those objects that could help them in understanding the underlying mechanism better We tested our algorithm on various datasets and present the results in this thesis Acknowledgement I would like to thank my advisor Prof Deborah Silver for her support and encouragement while writing this thesis Also I would like to thank my parents and family who provided strong educational foundation and supported me in all my academic pursuits I also acknowledge the help of VIZLAB at Rutgers Table of Contents ABSTRACT OFTHE THESIS eege Ee AEN dE EEN EELER dag 1 Ser FE EE 2 1 An Overview lde id TH TE E 5 2 Introduction to Scientific Visualization cinc rc 8 2 1 Underlying principles saartaa eea aaa ra pea aaa aro Aeae aaa aeaa aaa aaa a aa aaao t oaae eoi aan aak diaa 8 2 2 The Goal of scientific TEE E DT EE 8 2 3 Steps involved in scientific visualization s ssssssssssssususunseunnnnununnnunununnnnnnnnunununnnnnnnnannnnnanunnnnnnnnnnanannnnnnnnnna 8 2 4 Scientific Visualization to0 S cnica 9 2 5 Scientific Visualizat
89. s what Polydata plot does It uses traktable to assign correct colors to objects in a frame 5 3 3 Auto Generated Files VisIt provides with some tools to help developers add new plugins These tools generate a folder for the plugin with all the essential files needed by Visit in it Table 2 lists all auto files generated during the creation of all 4 Feature Tracking plugins using one of the tools xml2plugin Since FeatureTrak operator was modified from previous work 23 without any change to the operator attributes specified in xml file FeatureTrack xml was copied from previous work The steps involved in creating the plugin framework are listed below Plugin Names Auto Generated Files FeatureTrack TrackPoly PolyData Poly avt lt PluginName gt Filter C h v v v avt lt PluginName gt C h vV avt lt PluginName gt FileFormat C h avt lt PluginName gt Options C h avt lt PluginName gt Write C h Input LC bh java pyInput C h lt PluginName gt Attribute C h java py lt PluginName gt Attribute C h Qvis lt PluginName gt Window C h Qvis lt PluginName gt Window_moc LC h lt PluginName gt CommonPluginInfo C h lt PluginName gt EnginePluginInfo C h lt PluginName gt PluginInfo LC hi lt PluginName gt GUIPluginInfo LC hi lt SSIS S a S S S S S a lt S S S S S S S
90. sers by showing a mapping of topographical surfaces representing huge quantities of data in curve plot or scatter plots In Figure 6 we can see that Images annotation can be one of the visualization techniques to convey the results tenman e a b Figure 6 Examples of Visualization in formal sciences a Curve plot b Image annotations c Scatter plot Images source http en wikipedia org wiki Scientific_visualization Applied sciences Visualization tools are very useful for manufacturing and automobile industry These tools reveal a lot of information about the design of cars and aircrafts without actually manufacturing them thus saving a lot of money These tools are used to model cars study the aerodynamics of an aircraft and render traffic measurement in the city for city planners to come up with effective traffic management solutions a b o Figure 7 Examples of Visualization in Applied Sciences a Mesh plot of Porsche 911 model b Display plots of a dataset representing YF 17 jet aircraft c City Rendering results from rendering the description of building footprints in a city Image source http en wikipedia org wiki Scientific visualization 19 3 Feature Tracking Most complex simulations and observations generate data over a period of time Such time varying data have one or more coherent amorphous regions called Features 27 that might be of interest to the scientists or the users
91. set or a NULL object indicating that we should switch to Scalable Rendering mode 4 Subsequent actions like queries and picks cause the engine to connect new sinks to that AVT network 4 4 Plugin types VisIt supports development of custom plugins In VisIt plugins are divided into three categories plots operators and database readers and writers Plot A plot is a viewable object created from a database that can be displayed in a visualization window VisIt provides several standard plot types that allow you to visualize data in different ways The standard plots perform basic visualization operations like contouring pseudocoloring as well as more sophisticated operations like volume rendering All of VisIt s plots are plugins so you can add new plot types by writing your own plot plugins Operator An operator can be considered as a filter applied to a database variable before the compute engine uses that variable to generate a plot VisIt provides several standard operator types that allow various operations to be performed on plot data The standard operators perform data restriction operations like planar slicing spherical slicing and thresholding as well as more sophisticated operations like 36 peeling off mesh layers All of VisIt s operators are plugins and you can write your own operator plugins to extend VisIt in new ways Database VisIt can create visualizations from databases that are stored in many types of under
92. t Reset Max Active plots Hide Show Delete Draw E Pos Di Polntsize Point Type TrackPoly Dperatonetinbutee al Skew factor Opacity Smoothing Plots Operators Paus OpAtts Variables Sor ere size by var el Apply M operators selection to all plots Pointsize by var Point size pixels Path ERATED_TRACK FILES Enter the object SelectiveObjTracking v No Yes number s to be tracked ae ject Number to Track je Enter the opacity eet forrest of the objects in the frame L Al admin miracle E vist 1 12 0 Ip Window Figure 26 TrackPoly operator and PolyData plot attribute window The objects to be tracked is entered in textbox encircled in red 5 in this case and with the slider we can control the opacity for the other objects in the frame encircled in blue 76 3 The slider Opacity with default value 100 controls opacity for the objects in the list 4 Opacity of all other objects is controller by slider Opacity For Other Objects with a default value of 5 Then click Apply And if you press Draw only the objects in the list will be drawn as opaque and other objects becomes transparent Figure 21 displays the results for selectively tracking a single object object No 5 We have chosen opacity value as 5 for rest of the objects This is a default value and this can be set 0 to completely remove all the objects from back
93. tain consistency in our module we add 1 to whatever object number Vislt selects For e g if the object number of the object double clicked is 4 then we added 1 to it and display it as object 5 because our numbering of objects starts from object1 and not object0 5 6 Technical Challenges with Visit Unpredictable cycle time information One of the biggest hurdles towards development of custom s plugins for time varying datasets was the inability of VisIt to identify the current timestep The working of VisIt can be broadly described as follows e Data is read from the files via a database reader e Incase user wants to do some processing an operator is selected and the data passed on to plot for rendering on the screen e Now the plot can make any additional changes before displaying the data on the screen Colormaps are generated for each frame and these are applied to the data by PolyData plot If a colormaps are not applied to their corresponding frames visualization results go awry VisIt calls each frame as cycle and employs a guessing mechanism to guess the cycle or frame number If this guessed cycle number was used as the current frame and corresponding colormap chosen there always was an 60 error If guessed cycle number was off by one it could have been fixed easily but predicting cycle number was hard Even though GUI appeared to show correct cycle number plot was not getting the correct information Hence we devised a
94. tallation These plugins can be installed manually too You need to create 4 folders for the plugins two in operators one in plots and one in databases These folders should be names as that of the plugin The source code for these plugins is available at http miracle rutgers edu The steps for compiling and installing are similar to the ones mentioned for installation of FeatureTrack operator in Chapter 5 71 Feature Tracking amp Visualization User manual Feature Tracking amp Extraction Given a time varying 3D dataset the first step in the feature tracking process is to extract the objects in the individual frames and track them For this we need to apply FeatureTrack operator on the data Since Operator cannot work in unison and must always be accompanied by a plot we select pseudocolor plot for this purpose This step creates 5 files for each frame or timestep These are poly attr uocd trak colormap lt frame gt txt look at Appendix I for format of each file Also trakTable is created which tracks the interaction of objects from the first to the last frame or timestep FeatureTrack Attributes Figure 23 shows the attributes for FeatureTrack operator The description of the attributes for first 10 frames of a vorticity vorts dataset is as follows e inputBaseFilePath This is the path where the input files are located The input files are the files of bov vtk or any other file format that VisIt sup
95. that would be shown to the general public The amount and level of detail required in visualization is based on experience of the intended audience 2 3 Steps involved in scientific visualization At one level scientific visualization can be thought of analytically as a transfer function between numbers and images At another level visualization involves a barrage of procedures each of which 9 may influence the final outcome and the ability to convey meaningful information The process of visualization roughly consists of the following steps Data Filtering includes cleaning and processing of the data to yield meaningful results Examples would be removing noise replacing missing values clamping values in a certain range or producing new numeric forms leading to greater insights Representation Visual representation of information requires certain literacy on the part of the developer and the viewer 9 Beyond numerical representation of the output of the simulation it s advisable to give information about the simulation itself for e g the grid of the computation domain coordinate system scale information and resolution of computation The goal of visualization limits the medium of delivery which in turn puts constraints on the possible choices of representations So for example if the motion is an important aspect to show from the data then a medium that supports time varying imagery should be used Feedback It is a g
96. ting or picking objects with Mouse Click The first step is selection of either pick mode zone or node we select Zone Pick Mode in this case Then we have to select the objects We select two objects one is on the left and labeled as Object A the other one on the right is Labeled B 78 All this does is automatically append the object number to the list To see the results click Apply in PolyData plot attribute window From here its same as Selective Feature Tracking and you get results as shown in figure 21 Only difference being the Object Number selected GB Applications Places system DS Qi o Visit 1 12 0 Eile Controls Options Windows vorts5 poly cy P Lighting untah eh m ReOpen Use min oe 20 Ai A Scale Min Active window Max Active plots Hide Point size Point Type Skew factor Opacity Smoothing Plots Operators Apply Hoperators A Point size by var enabled Point size by var Point size pixels Path SelectiveObjTracking Number to Track o x Window 1 Linear v Log v Skew e 1 0 05 0 1 Hj 0 default f 2 NERATED_TRACK_FILES Yes v No D Pick AE vorts1 poly timestep 0 mesh Point lt 17 4894 71 2057 94 2827 gt Zone 13493 Incident Nodes 6844 6843 6950 var lt nodal gt Max Tabs pe t QA E ot w
97. tion This hampers the development of new features or plugins in VisIt A lot of time and energy is spent in understanding the design so this thesis also aims to document the design decisions that could be useful to other developers The organization of the thesis is as follows Chapter 2 provides an introduction to Scientific Visualization and describes some software packages like AVS ParaView and VisIt Chapter 3 describes Feature Tracking process and various software implementations that exist in Vizlab Chapter 4 is about Visit We discuss the design working and procedure for creating new plugins in Visit In Chapter 5 we talk about the functionalities that have to be added to VisIt design of Feature Tracking amp Visualization plugins new methods added modifications made to the existing code and the challenges faced during this process of designing new plugins The algorithm was tested on different datasets and results presented in Chapter 6 Finally thesis concludes with reiterating the usefulness of Feature Tracking in VisIt and its benefits as a scientific visualization tool We provided a detailed user manual about the installation procedure and information on using the plugins in appendix 2 Introduction to Scientific Visualization As stated in 8 Scientific visualization is an interdisciplinary branch of science primarily concerned with the visualization of three dimensional phenomena architectural meteorological medical
98. tructing an AVT network There is no communication between different components of Vislt The viewer just sits with the information and does nothing with it 3 Clicking on draw As the user clicks on Draw the AVT network is executed and the following steps are executed during this process e Preparing for scalable rendering e JInnon scalable rendering the resulting surface is transferred to the viewer and rendered locally The rendering is done using an avtPlot s mapper module being called in the context of a VisWindow s visualization window e In scalable rendering the surface is rendered in parallel and the engine transfers an image back to the viewer e Stating which file to use as the source RPCExecutor lt ReadRPC gt Execute from Executors hh is called This method calls NetworkManager StartNetwork and the following actions take place 1 The avtDatabase is identified it was already created during an OpenDatabaseRPC 2 An avtTerminatingSource is gotten from the avtDatabase 34 3 An avtExpressionEvaluatorFilter is added to the pipeline at the top 4 The avtSILRestriction is registered e Setting up the operators 1 RPCExecutor lt PrepareOperatorRPC gt Execute is called This must be called first to instantiate the correct type of attributes so that the subsequent call to Add Operator will be able to load the attribute values into this instance 2 RPCExecutor lt AddOperatorRPC gt Execute is called Th
99. ttribute of the plot This slider has a default value of 5 that be changed When user clicks on Apply a list of non opaque objects to be tracked in subsequent frames is generated This list of non opaque objects depends on the initial list provided by the user as an input and the trakTable file Any sort of processing or action to be taken after a user clicks on Apply and before rendering the results can can be obtained by modifying the methods in QvisPolyDataPlotWindow class Any series of events or processing that is based on action buttons on the GUI of the plugin requires modifications to methods in QVis lt Pluginname gt Window class To generate the list of non opaque objects new methods are added to QvisPolyDataPlotWindow class Figure 13 These methods require the name of the current dataset to open correct trakTable file Information such as the name of the dataset number of frames etc comes from CurrentFile txt see Appendix I for the format CurrentFile txt is described later on in the section of technical challenges within Visit 57 5 3 2 1 New Methods in PolyData Plot The signature of the new methods added and their behavior is described below 1 void CreateOpacityTable string string This method is called iteratively for the frames in the dataset The input for this method is the object s to track and the location where the file has to be created With the help of helper methods described below it tracks the
100. tual avtDataObject_p virtual avtDataObject p virtual void virtual void SetAtts const AttributeGroup GetMapper void ApplyOperators avtDataObject_p ApplyRenderingTransformation avtDataObject_p CustomizeBehavior void CustomizeMapper avtDataObjectInformation amp All these methods part of the class avt lt plot name gt and hence would be defined and declared in avt lt plot name gt C and avt lt plot name gt h Depending on the aim of the plugin different methods have to be modified 39 While adding an operator the important method is vtkDataset ExecuteData vtkDataSet int std string This method is part of class avt lt operator name gt Filter is declared and defined in avt lt plot name gt h and avt lt plot name gt C Any processing or filtering on the dataset can be introduced by adding some lines of code in this method While writing a database writer there are four basic methods which must be implemented although any of these methods can be no ops i e you can leave them empty with just The specifics on how these methods are called is mentioned in src avt Pipeline Sinks avtDatabaseWriter C and the signature of these methods are void OpenFile const std string amp int void WriteHeader const avtDatabaseMetaData std vectorsstd string gt amp std vector lt std string gt amp std vector lt std string gt amp void WriteChunk vtkDataSet int void Clos
101. twork building by color coding the ports Input and output ports of similar color are connected indicating the flow of the data in the network Since data is primary to perform any processing the first module is always be for reading the data and the last module is a viewer module to display the results Figure 6 shows a simple network in AVS to read field files downsize the data according to some criteria and produces the orthoslices for the volume We have modules for bounds and Axis3D in the 12 resulting image we see a bounding box and axis for the volume Any change in data for any of the module causes entire network to be executed again to represent the change in visualization yA A a a Figure 2 AVS Network Image source http help avs com express doc help_722 index htm 2 5 2 ParaView ParaView is an open source multi platform data analysis and visualization software package With ParaView users can quickly build visualizations to analyze their data using qualitative and quantitative techniques The data exploration can be done interactively in 3D or programmatically using ParaView s batch processing capabilities It has been successfully tested on Windows Mac OS X Linux IBM Blue Gene Cray Xt3 and various Unix workstations clusters and supercomputers Under the hood ParaView uses the Visualization Toolkit VTK as the data processing and rendering engine and has a user interface written using Qt Some
102. txt format in Appendix I to keep track of current frame In this thesis we modify the format of the curpoly txt file from previous work 23 45 5 3 2 2 TrackPoly operator This plugin reads curpoly txt and replaces the last line in the file indicating the frame number with the current frame number 5 3 2 3 PolyData plot A plot should be able to show the results on screen PolyData plot aims to visualize the objects listed in a poly file by displaying them on the screen TrakTable based coloring scheme allows us to differentiate objects in a frame and track them over a period of time This plot tracks the objects and assigns them colors accordingly One may ask why do we need to have a separate plot for this purpose Can t plots like pseudocolor perform this task The answer is NO because none of the other plots use tracking information for coloring so we have to design new plot for this purpose Pseudocolor plot assign colors based on iso surface values All distinct iso surfaces for a timestep would be allocated different colors So lets say there are three different iso surfaces so we assign colors like red blue and green Now in the next frame one of the objects splits into 2 objects making a total of 4 objects for the next frame Now there are 4 distinct iso surfaces and they all should get different colors so the notion of tracking is gone The only way this can be achieved is by using trakTable information and that
103. vent querying Using the accumulated tracking information we can also perform additional visualization steps like event querying which involves gathering information leading to a certain event of interest or present a new visualization using the data metadata collected One example of this is volume rendering of an individual feature 22 Different Feature Tracking algorithms at Vizlab are classified as 3 2 1 Overlap based Feature Tracking Algorithms These define five classes of interactions that are used by many other algorithms 34 These interaction classes are given below and are illustrated in Figure 13 1 Continuation An object continues from time bo Ga with possible rotation translation or deformation Its size may remain the same grow or shrink 2 Creation New objects appear 3 Dissipation Objects disappear Dissipation generally occurs when regions fall below the specified threshold value 4 Splitting Bifurcation An object splits into two or more objects 5 Merging Amalgamation Two or more objects merge Es ez t 1 ea amalgamation t 2 Y dissipation ES Figure 10 Tracking Interactions Continuation creation dissipation bifurcation and merging Image source Lian s Thesis An overlapping based Feature Tracking can be based on Octree datastructure or Linked List based The octree based algorithm works in two phases 23 1 VO test
104. view its evolution separately in a different window e Attribute analysis and Printing The software computes various attributes like volume mass centroid etc that can be printed on the screen by picking a particular object from a time step in the enhanced surface animation window 27 e Graph Plotting The interface also has a window where the user can view how some frame attributes like number of objects etc vary over the time duration of the tracking Storing of Feature tracking results All attributes for individual objects as well as for all time steps are stored in files in a pre defined directory under the users run path The files also include a record of the events which occur in the life cycle of an object e g splitting or dissipation 3 4 2 Distributed Feature Tracking Implementation Distributed Feature Tracking was implemented as a standalone C application 38 Gives implementation details of Distributed Feature Tracking code Given a huge dataset the code would work in parallel mode distribute the task among a group of processors to extract features from all the frames and store the tracking information in a trakTable file The code is organized in 4 separate directories Pinakin s thesis These directories are objseg directory During the feature extraction step each processor loads a local portion of the dataset At the end of this step each processor generates poly file oct file object attri
105. when Poly database reader reads a poly file Once this file is created it s modified by the TrackPoly operator to indicate the current frame number Other than the frame number this file also records the name and the number of objects in the frame This file is just to double check whether the current frame number is conveyed properly or not as the cycle number in CurrentFile txt also represents this information This way we can be sure that PolyData plot gets the correct colormap information Sample curpoly txt vortsl poly Filename 26 Number of object 1 Frame number OpacityTable txt This file is generated only if the selective tracking option is enabled There is only one file for the entire dataset This file has information about the non opaque objects in a frame and for every frame this information gets appended to the already existing file This is basically the tracking information for the object s mentioned in starting frame Sample OpacityTable txt file NonOpaqueObject_Frame 1 4 6 NonOpaqueObject_Frame 2 3 8 NonOpaqueObject_Frame 1 3 4 87 88 Appendix Il Data structures in Scientific Visualization A big hurdle in using any visualization system can be getting the data into the package The form of the geometry in the problem frequently dictates the form of the data e g a spherically symmetric problem results in spherical polar coordinates in the data set Each visualization
106. workaround that did not rely VisIt guessing mechanism This is described in next section Workaround VisIt automatically groups similar files in a folder as a database This can be seen in Figure 16 where all the vorts poly files in that location are grouped together as vorts database Double clicking any of the files opens the database and Visit gets a corresponding cycle number by employing a guessing mechanism All this functionality is provided by VisIt s GUI module One of the classes associated with VisIt s GUI module is QvisFilePanel QvisFilePanel class has a method called UpdateFileSelection that monitors the files in the database and reports whenever a new file is selected A user can load any file in the list by double clicking on the file or using a animation slider to load the required file This action would call the UpdateFileSelection method While making the database of the files VisIt gather all the metadata information associated with the files in the database From this metadata information we extract the location of the file also called the Path the number of files frames in the database the starting amp the ending frame or timestep number the current file frame and the name of the database We write this information to a text file called CurrentFile txt see appendix I for the format of this file Now VisIt waits for user to add operator or plots and execute the network The reader loads the data only
107. written to trakTable file This part was implemented as plugin in VisIt that could be easily installed by any other user Another plugin was designed for visualizing the features extracted from the time varying datasets However these plugins were found to be incomplete and incompatible with newer versions of Vislt In this thesis we use the framework from 23 and make modifications to it The plugins names are the same as in previous work but the functionality of these plugins have been modified 29 4 How Visit Works The basic design of VisIt can be thought of as a client server model 14 The client server aspect allows for effective visualization in a remote setting VisIt s architecture allows for parallelization of the server task of one processor is shared by a group of processors thereby processing of large datasets quickly and interactively VisIt has been used to visualize many large data sets including a two hundred and sixteen billion data point structured grid a one billion point particle simulation and curvilinear unstructured and AMR meshes with hundreds of millions to billions of elements VisIt follows a data flow network paradigm where interoperable modules are connected to perform custom analysis The modules come from Vislt s five primary user interface abstractions and there are many examples of each In VislIt there are e twenty one plots ways to render data e forty two operators ways to manipulate d

5. Feature Tracking & Visualization in VisIt

Contents

Download Pdf Manuals

Related Search

Related Contents

5. Feature Tracking &amp; Visualization in VisIt

Contents

Download Pdf Manuals

Related Search

Related Contents

5. Feature Tracking & Visualization in VisIt