Home

fire alert system using shape and color analysis through image

Contents

1. close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG M text 0 1 MeanRGB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox 55 bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 eplot bo 1 bc 2 mt a text bc 1 15 bc 2 Y num2str round bc 2 set a FontName 12 Color yellow drawnow end if frameCounts 20 frameCounts else frameCounts 0 break end fwrite arduino angle2 pause 0 02 Arial strcat X num2str round bc 1 FontWeight bold FontSize frameCounts 1 Write to arduino angle 45 degrees videoFram origVid videoFrame redVid videoFrame greenVid videoFrame blueVid videoFrame ex Red imsubtract redVid ex Red medfilt2 ex Red out noise ex Red im2bw ex Red 0 18 5 image into a binary image step videoDevReader rgb2gray videoFrame
2. 3 3 Use a median filter to filter Convert the resulting grayscale Remove all those pixels less x Red bwareaopen ex Red 300 3 than 300px ex Green x Green filter out noise ex Green image into a binary image imsubtract greenVid medfilt2 ex Green im2bw ex Green 0 18 rgb2gray videoFrame 3 3 Use a median filter to Convert the resulting grayscale n 300 Remove all those pixels x Gr less than 300px n bwareaopen ex Gr 56 ex Blue imsubtract blueVid rgb2gray videoFrame ex Blue filter out noise medfilt2 ex Blue 3 3 Use a median filter to ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oo Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Bl
3. the image oe Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel T subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROC Lr r 64 ESSING cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION 2f SR ST text 0 0 sprintf capsT Determine the camera devicelID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end
4. Dbg 2 T M a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end 84 if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end end end 9 disp Program End Message sent to command window return 85
5. FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end fwrite arduino angle4 Write to arduino angle 135 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px 63 ex Green imsubtract greenVid rgb2gray videoFrame x Green medfilt2 ex Green 3 3 Use a median filter to filter out noise ex Green image into a binary image x Green bwareaopen ex Green 300 Remove al less than 300px oO X w G 0 ll im2bw ex Green 0 18 Convert the resulting grayscale l those pixels imsubtract blueVid rgb2gray videoFrame ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise 9 ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in
6. Procedures in conducting the response of android mobile application testing are 1 Set up the fire alert system 2 Burn paper or any material 39 3 Wait for the system to detect fire 4 If the system detects a fire start the stop watch 5 Stop the stop watch when the notification pops out in the android phone 6 Record the result time Fire Detected Burning Paper Burning Rubber Burning Plastic Burning Wood 40 No Fire Detected t Lighter Cigarette Gas Stove Burner 41 Chapter 4 CONCLUSION The study Fire alert system based on shape and color analysis through image processing via Android mobile application focuses on implementing a fire alert system that will defect early fire occurrence and alert users through the mobile By performing trials of tests the researchers can say that the system is able to detect fire with the range of 5 seconds to 20 seconds from the start of the flame The time of detection may vary depending on the quality of the camera The system created an android mobile application that will automatically inform the user if there is a fire detected Time interval from fire detection to fire alert android mobile application depends on the network provider With the use of image processing the researchers created a system that can detect fire on its early stage By integrating image processing using MATLAB application on the camera with the help of shape and color analysis this
7. component is a different point through which the system can enter an application and instantiate component object independently Jianye Liu and Jiankun Yu 2011 According to the article Context aware User Interface Framework for Mobile Applications mobile devices today are very heterogeneous using different platforms and featuring a multitude of different display resolutions Therefore it is often necessary to adapt an application to multiple devices at development time taking into account the different display and input capabilities and developing different implementations for different devices The number of available platform configurations such as Java Platform Micro Edition Java ME also referred as J2ME CDC or CLDC further increases the number of different implementation needed for a portable application It is known that a user interface should be able to adapt to the current situation of the user to better suit its needs and desires thus improves usability The XML User Interface Language XUL is an XML based mark up language for the description of user interfaces It is mainly used by Mozilla Foundation in their products like the Firefox browser or Thunderbird mail client but is increasingly popular in the area of web applications XUL is used by several other projects for mobile devices to specify user interfaces and transforms the XUL to device specific HTML or WAP pages In this paper the researchers will introduce a fra
8. off cla MeanRGB sprintf R 2f G 2f B 2 f MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 f MR MG MB text 0 1 MeanRGB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 plot be l be 2 mt a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end fwrite arduino angle4 Write to arduino angle 135 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame ex Red out noise medfilt2 ex Red 3 3 Use a median filter to filter ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px ex Green imsubtract greenVid rgb2gray videoFrame x Green medfilt2 ex Green 3 3 Use a median filter to filter out noise o ex Green im2bw ex Gre
9. to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION Sedi SR ST text 0 0 sprintf capsT Determine the camera deviceID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaghwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided 9 The following sections takes video according to the time specified by the user k 1 while 1 72 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my _ log refnum refnum 1 if k loopnumber de
10. 0 sprintf capsT Determine the camera devicelID imagreset infom imaqhwinfo for k 1l length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DevicelIDs 68 Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided 9 The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k gt loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB 69 end else subplot 2 4 5 8 Visible
11. 2 2 Make sure you have internet connection on your phone 3 Video feedback from the camera cannot be seen in the android phone 3 1 Check if the camera is on and is connected to the router 3 2 Make sure you have internet connection in your phone 3 3 Check if the IP address of the camera is correct Error Definitions 1 No internet connection 2 Too much process on the computer 3 Human error 48 APPENDIX B Pictures of Prototype Registration failed Please try again Messages Q Goi ile Android Application Notification Setting up Fire 49 R 041G 0328B 0 30 FIRE DETECTED PROCESSING E Fire Detected 50 Image Processing Camera IP camera and router 51 APPENDIX C Program Listing function FireAlertSystem clc Clear command window clear all Get rid of variables from prior run of this m file disp Program Start Message sent to command window workspace Show panel with all variables delete instrfindall Find visible and hidden instrument objects imaqreset Disconnect and delete all image acquisition objects imagreset deletes any image acquisition objects that exist in memory and unloads all adaptors loaded by the toolbox As a result the image acquisition hardware is reset oo imaqreset is the image acquisition command that returns MATLAB to the known state of having no image acquisition objects and no loaded
12. Bachelor of Science in Computer Engineering dusx S PaL Felicito S Caluyo Dean School of EECE ACKNOWLEDGEMENT The designers would like to thank the following people for their effort in making this thesis possible First of all the designers would like to thank God for the knowledge and wisdom and the strength to be able to finish this thesis To Engr Analyn N Yumang for providing guidance in developing the program as well as in making the documentation for this thesis To our beloved panellists Engr Dionis Padilla Engr Joshua B Cuesta and Engr Jose B Lazaro Jr for assistance in improving this paper To Engr Voltaire De Leon and Engr Ayra Panganiban for the time and guidance extended to the researchers TABLE OF CONTENTS TITLE PAGE APPROVAL SHEET ACKNOWLEDGEMENT TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES ABSTRACT CHAPTER 1 INTRODUCTION CHAPTER 2 REVIEW OF RELATED LITERATURES AND RELATED STUDIES Android Applications IP Camera Related Studies Image Processing based Fire Detection System A New Image Based Real Time Flame Detection Method An Early Fire Detection Method Based on Image Processing Flame Detection Method in Video Using Covariance Descriptors Fast and Efficient Method for Fire Detection Using Image Processing Implementation of Object Tracking and Motion Sensors CHAPTER 3 FIRE ALERT SYSTEM USING SHAPE AND COLOR ANALYSIS THROUGH IMAGE PROCESSING VIA MOBILE APPLICATION Abs
13. There are fire detection systems that use smoke sensors which require a certain concentration of the smoke before it can detect fire but can lead to false detection such as a cigarette smoke The common problem where fire starts at home is caused primarily by homeowners 19 being careless when it comes to dealing with potential hazardous situations Another cause usually deals with electricity where people forget to unplugged electric appliances The problem in the existing studies is accuracy in terms of time on when will the device detect the flame The system that will be created is very helpful for the safety of everyone If the system will be installed in houses the families safety will be at hand and if the system will be installed in establishments the owners ensured customers safety Nowadays the Earth s average temperature increases due to the effect of greenhouse gases which is why fire and floods happen unexpectedly In 2012 during the summer time here in the Philippines temperature reaches 36 degree celsius and strong typhoons enter the Philippines area of responsibility Because of fire and flood huge damages happen and worst people die In fact these circumstances can be minimized if people will be alarmed and can call for help immediately The system that will be created will lessen these huge damages at the same time will alert the owner immediately if the IP Internet Protocol camera saw an occurrence of fire or smo
14. VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k gt loopnumber delete vid clear vid break end 58 end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 f MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 Splot be 1 5Db9 2 Tmt a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end 59 fwrite arduino angle3 Write to arduino angle 90 degrees pause 0 02 videoFrame step videoDevReader or
15. cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION Suk SR ST text 0 0 sprintf capsT o Determine the camera deviceID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided o The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI 54 vid YUY2 320x240 vid LoggingMode vid DiskLogger vid TriggerRepeat set vid FramesPerTri start vid videoinput adaptorname deviceid disk amp memory aviobj int igger 1500
16. end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided 9 The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid 65 close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end subplot 2 4 1 for object 1 length
17. ex Red 300 Remove all those pixels less than 300px ex Green imsubtract greenVid rgb2gray videoFrame x Green medfilt2 ex Green 3 3 Use a median filter to filter out noise ex Green im2bw ex Green 0 18 Convert the resulting grayscale image into a binary image x Green bwareaopen ex Green 300 Remove all those pixels less than 300px 74 ex Blue imsubtract blueVid rgb2gray videoFrame ex Blue filter out noise medfilt2 ex Blue 3 3 Use a median filter to 9 ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oo Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 M
18. fire text message the application will convert this text message to open the fire alert android mobile application and the user can view what is happening in the scenario The live streaming speed depends on the service provider The researchers are able to develop a fire detection system using image processing and creat an algorithm for fire detection using shape and color analysis With the algorithm created a fire detection system is more accurate than the past researches and systems developed 43 Chapter 5 RECOMMENDATION Based from the limitations of the fire alert system and the results gathered the system can be further improved by the following recommendations First is increasing the fire detection accuracy of the image processing program to distinguish between the background color of the room and fire color if both are of the same color thus eliminating the remaining false alarm of the system Second recommendation is increasing the camera s range to suit bigger rooms which also increases the fire detection accuracy Also to further improve the system s capability in fire detection future works can be integrated to the temperature sensor that can detect room temperature This can help the user in assessing the severity of the situation For the android mobile application the option to record the live video stream can be added because this can help the authorities in conducting investigations on the cause of fire 44 APP
19. from the image To get rid of these spurious fire like regions the image difference method and the invented color masking technique are applied Finally a simple method is devised to estimate the burning degree of fire flames so that users could be informed with a proper warning alarm An Early Fire Detection Method Based on Image Processing The paper presents an early fire detection method based on video processing The basic idea of the fire detection is to adopt an RGB red green blue model based chromatic and disorder measurement for extracting fire pixels and smoke pixels The decision function of fire pixel is mainly deduced by the intensity and saturation of R component In addition to using chromatics dynamic features are adopted to distinguish other fire aliases These fire dynamics include sudden 13 movements of flames changeable shapes growing rate and oscillation or vibrations in the infrared response For improving the reliability of detection it utilizes both the disorder characteristic of flames and the growth of fire pixels to check if it is a real fire Since the shape of flames changes with time due to air flowing the size of the fire s area in an image also changes the flame always has a growth feature The disorder of fires can be measured with the pixel quantity of flame difference between two consecutive images Flame Detection Method in Video Using Covariance Descriptors This paper presents a video f
20. has been proven that the system is more accurate compared to other research studies and other fire or smoke detectors With the use of the created algorithm the system can detect or distinguish the difference between a fire and an object that has the same shade color of a fire The only thing where one can get false alarms is when the background color is the same as the color of the fire Based from the results gathered there are factors that play an important role in getting the output First is the distance of the camera from the fire and second is the camera on how it can detect the fire quickly The web camera is responsible for 42 image processing which is through the use of MATLAB programming The web camera s range is 10m or 30ft but is limited to the system design to 5m or 15ft because of the low quality of the web camera and to make the system more accurate When the researchers are developing the system it was found out that the web camera couldn t save video clip while the user is viewing the scenario that is why the researchers use two 2 cameras The web camera is used in image processing and saved video clip while the IP camera is used in viewing the scenario As the researchers finish the system all of the objectives successfully met They are able to create a android mobile application that will automatically inform and alert the user The application will automatically receive an SMS alert message when the system send a Fire or
21. image acquisition adaptors You can use imaqreset to force the toolbox to search for new hardware that might have been installed while MATLAB was running Note that imagreset should not be called from any of the callbacks of a videoinput object such as the StartFcn or FramesAcquiredFcn set gcf Position get 0 ScreenSize set gcf name Fire Alert System Using Shape and Color Analysis Through Image Processing via Mobile Application numbertitle off videoDevReader imaq VideoDevice winvideo 1 YUY2 320x240 set videoDevReader ReturnedColorSpace rgb subplot 2 4 5 8 Visible off capsT sprintf INITIALIZATION COMPLETE cla text 0 1 capsT drawnow while 1 1 while 1 1 fwrite arduino anglel SWrite to arduino angle 0 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame 52 ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px ex Green imsubtract greenVid rgb2gray videoFrame x Green medfilt2 ex Green 3 3 Use a median filte
22. imsubtract redVid rgb2gray videoFrame ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px ex Green x Green medfilt2 ex Green 3 3 filter out noise o image into a binary image 9 less than 300px x Green bwareaopen ex Green 300 imsubtract greenVid rgb2gray videoFrame Use a median filter to ex Green im2bw ex Green 0 18 Convert the resulting grayscale Remove all those pixels ex Blue imsubtract blueVid rgb2gray videoFrame ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oo Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox MR mean mean redVid 78 Centroid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel ES subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red ti
23. lighting conditions may adversely affect the saturation values of flames resulting in similar R G and B values which may cause non flame pixels to be considered as flame colored Therefore saturation values of the pixels should also be over some threshold value In the third condition S gt 255 R S R where S is the value of saturation when the value of R channel is R If all of the three conditions are satisfied for a pixel then that pixel is considered as a fire colored pixel The values of both R and are typically 0 5 and 0 11 respectively 32 Testing and Statistical Treatment Material Used Correct Detection 9 o False Detection 9 o Paper 100 0 Rubber 97 6 2 4 Plastic 98 5 1 5 Table 3 1 Performance and Evaluations Table 3 1 presents the performance and evaluation of the fire alert system The correct detection refers to the success rate of the system in detecting fire while the false detection refers to the percentage of unsuccessful detection of the system Using different materials the system is tested using 25 samples of each material The results obtained from the testing are presented in the table Procedures in conducting the performance and evaluation testing are 1 Set up the fire alert system 2 Burn the said material 3 Wait until the system detects the fire If the fire becomes bigger and the system doesn t detect this is considered as a false detection 4 Wait for the android
24. since the Android mobile platform s middleware is written in Java and C C The figure below discusses the process flow of the mobile application The mobile application initially runs as a background application on the mobile phone Is the Application ON Is there YES Fire Connect to IP Show Live Detected Camera Video Figure 3 4 Android Mobile Application Algorithm 27 Figure 3 4 discusses the android mobile application algorithm once the system detects a fire it will send a message that will be connected to the mobile phone The mobile application connected to the IP camera will begin the live video streaming of the area covered 28 Extract image frames from video Segment moving regions by image differencing Color Analysis Shape Analysis Fire Detected Send Fire Alarm Figure 3 5 Shape and Color Analysis Algorithm The proposed fire detection algorithm can be concluded in Figure 3 5 Firstly the moving regions are segmented from the captured image frames thus used as candidates for checking if it is a fire or not The shape created by segmenting these moving regions would be verified by testing each pixel using the 29 color analysis By using the chromatic color model each pixel would be tested if it is a fire or not It was analyzed that the hue of the fire colored pixels was in the range of 0 and 60 The RGB domain equivalent of this condition is R 2 G g
25. stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 Splot bo 1 Dbg 2 T M a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end 66 if frameCounts 20 frameCounts else frameCounts 0 break end fwrite arduino angle5 pause 0 02 videoFram frameCounts 1 Write to arduino angle 180 degrees origVid videoFrame redVid videoFrame greenVid videoFrame blueVid videoFrame step videoDevReader t bL rgb2gray videoFrame 3 3 Use a median filter to filter Convert the resulting grayscale 5 Remove all those pixels less ex Red imsubtract redVid ex Red medfilt2 ex Red out noise ex Red im2bw ex Red 0 18 image into a binary image x Red bwareaopen ex Red 300 3 than 300px ex Green x Green filter out noise ex Green image into a binary image imsubtract greenVid medfilt2 ex Green im2bw ex Green 0 18 rgb2gray videoFrame 3 3 Use a median filter to Convert the resulting grayscale o Remove all those pixels x_Gr less than 300px ex Blue ex Blue filter out noise ex Blue image into a binary image n bwareaopen ex Green 300 imsub
26. 1 else frameCounts 0 break end fwrite arduino anglel Write to arduino angle 0 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px 81 ex Green imsubtract greenVid rgb2gray videoFrame x Green medfilt2 ex Green 3 3 Use a median filter to filter out noise ex Green image into a binary image im2bw ex Green 0 18 Convert the resulting grayscale x Green bwareaopen ex Green 300 Remove all those pixels less than 300px oO X w G 0 ll imsubtract blueVid rgb2gray videoFrame ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oe Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw Bou
27. CT Fire is a common safety problem both in the industry and household environments This paper presents the research and implementation of a fire detector through covariance matrix image processing and alarm device with mobile phone application The concept of covariance is used together with Shape analysis algorithm and color analysis algorithm The device can detect fire by its color knowing that there is an enrolment made for color on our web server A mobile phone application is created to serve as the receiving end of the device s alarm A moving camera is also utilized for live video streaming of a certain area The device will detect if there is a fire in a given area and then an alarm will be sent to the user s Android mobile phone From the mobile phone application the user will be able to view the happenings in the area and will have an option to call the fire station Keywords image processing fire web server android mobile application viii Chapter 1 INTRODUCTION Recently terrible disasters such as tsunami flood avalanche fire and explosion have caused many casualties and damages to property Therefore efficient detection for early alarming systems has to be used and has been an important research area in the last few years Researches on the automatic fire detection in particular has been widely studied Sensor based fire detection is being used due to its characteristics of low cost and simple set up However these dev
28. ENDICES 45 APPENDIX A Operation s Manual System Requirements These are the requirements necessary to operate the device It should be a laptop or a computer with the following features 1 Windows XP Service Pack 3 or newer versions 2 Intel Pentium 4 or higher 3 512 MB RAM minimum 1 GB RAM recommended 4 Atleast 625 MB Hard Disk space 5 Graphics 16 24 or 32 bit OpenGL capable graphics adapter 6 USB Port 7 Internet Connection Installation Procedure 1 Install the MATLAB Software in the PC or laptop 2 Install the application in the android phone 3 Connect the device to a USB port of a laptop or PC 4 Connect the camera to the device 46 5 Mount the camera on a solid level surface 6 Connect the device to its power source 7 Configure the router and the IP address of the camera User s Manual PC or Laptop 1 Run the MATLAB software installed in the PC or laptop 2 Connect to the internet 3 Run the program in MATLAB Android Phone 1 Connect to the internet 2 Run application in phone Troubleshooting Guides and Procedures 1 No video is shown in the output of program 1 1 Check the connection of the device to the PC 1 2 Make sure the device is connected to its power source 1 3 Adjustthe position of the camera 47 2 Unable to receive alert on the android phone 2 1 Check if the application is running in background on the phone If not restart the application
29. FIRE ALERT SYSTEM USING SHAPE AND COLOR ANALYSIS THROUGH IMAGE PROCESSING VIA MOBILE APPLICATION By Divina A Chua Carla Louie H Leandicho Leo Angelo C Magtibay Jerome T Ortiz A Thesis Report Submitted to the School of Electrical Engineering Electronics Engineering and Computer Engineering in Partial Fulfillment of the Requirements for the Degree Bachelor of Science in Computer Engineering Mapua Institute of Technology March 2013 Approval Sheet Mapua Institute of Technology School of EECE Thi is to certify that I have supervised the preparation of and read the design report prepared by Divina A Chua Carla Louie H Leandicho Leo Angelo C Magtibay Jerome T Ortz entitled FIRE ALERT SYSTEM USING SHAPE AND COLOR ANALYSIS THROUGH IMAGE PROCESSING VIA MOBILE APPLICATION and that the said report has been submitted for final examination by the Oral Examination Committee As members of the Oral Examination Committee we certify that we have examined this design report presented before the committee on March 2013 and hereby recommended that t be accepted in fulfilment of the design requirements for the degree in Bachelor of Science in Computer Engineering se B Lazaro Jr Panel Member 7 Padilla Committee Chairman This design report is hereby approved and accepted by the School of Electrical Engineering Electronics Engineering and Computer Engineering in partial fulfilment of the requirements for the degree in
30. application This study aims to develop an algorithm for fire detection using shape and color analysis The specific objectives of this research are as follow a To create a mobile application that would alert and automatically inform the user if there is a fire detected b To develop fire detection system using image processing c To develop a more accurate algorithm that is early fire detection The study impacts on the safety of the people as well as contribute greatly in the safety field through the convenience of an alert system with the use of a mobile application The research and the proposed algorithm are suited to the advancement of technology where people nowadays are seeking portability of things The portability of the device will be evident in such a way that the user will be alerted from the mobile phone theoretically anywhere provided that the mobile phone is powered on The mobile application can be used instantly once the web server distinguish that the captured image of fire transmit a signal Lastly this research will contribute in the future studies of wireless technology which is now one of the most widely used technology in transferring data or information The research also focuses on house monitoring of fire through image processing It alerts the user if there is a fire occurring in the said location The study covers the use of co variance based on fire color detection and shape analysis to determine a fire thro
31. camera to automate its viewing perspective The camera can scan the area pivoting from left to right and vice versa The camera can also be manually operated by the user using two push buttons to direct the camera to the left or to the right The use of sensors instead of multiple cameras have made it possible to monitor a large area while minimizing the use of cameras 16 BET aus EN Vo VT Ta To E Servo er als gee snapshot rr email Figure 2 4 Block Diagram Implementation of Object Tracking and Motion Sensors for Home and Office Based WIFI Monitoring System Chapter 3 FIRE ALERT SYSTEM USING SHAPE AND COLOR ANALYSIS THROUGH IMAGE PROCESSING VIA ANDROID MOBILE APPLICATION Abstract The study presents the research and implementation of a fire detection system through covariance matrix image processing and an alarm system through the mobile phone application The system extracts image frames from the video for shape and color analysis and segments the moving regions by image differencing through the concept of co variance used together with shape and color analysis A mobile phone application is created to serve as the receiving end of the system which moving camera is also utilized for analysing the environment of a certain area for flame colored objects The device will detect if there is a fire in a given area and then a message will be sent to the Android mobile ph
32. curity memory management file system management process management network group Driver Model etc xFace is composed of xFace engine and platform adaptation layer also known as porting layer xFace engine is mainly responsible for page layout script analysis logic and other functions the porting layer is mainly responsible for abstracting the operating system related functions into the system independent in order to make xFace run of different platform xFace is developed by C C but the Android application is developed by Java language which allows us to have to use JNI Java Native Interface which is provided by JAVA to allow JAVA and the C C interface to call each other Yonghong Wu Jianchao Luo Lei Luo 2010 Application xFace Core C C Porting Layer ndroid Android SDK NDK Android Linux Libraries Libraries JAVA C C Figure 2 2 xFace Porting Layer IP Camera An Internet protocol camera or IP camera is a type of digital video camera commonly employed for surveillance which unlike analog closed circuit television CCTV cameras can send and receive data via computer network and the internet Although most cameras that do this are webcams the term IP Camera is usually applied only to those used for surveillance Based on the article Cost Effective IP Camera for Video Surveillance IP cameras are the key devices for video surveillance system The IP camera can compress and deliver live vide
33. e maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end 62 else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R text 0 1 MeanRGB oe oe 2 G oe 2 B 2 MR MG MB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 plot bc 1 bc 2 m a text bc 1 415 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial
34. en 0 18 Convert the resulting grayscale image into a binary image x Green bwareaopen ex Green 300 Remove all those pixels less than 300px ex Blue imsubtract blueVid rgb2gray videoFrame ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise 9 ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oe Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow 71 subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 o Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvimage rgb2hsv videoFrame Convert the imag
35. for the safety of everyone If the system will be installed in houses the families safety will be at hand and if the system will be installed in establishments the owners ensured customers safety Nowadays the Earth s average temperature increases due to the effect of greenhouse gases which is why fire and floods happen unexpectedly In 2012 during the summer time here in the Philippines temperature reaches 36 degree celsius and strong typhoons enter the Philippines area of responsibility Because of fire and flood huge damages happen and worst people die In fact these circumstances can be minimized if people will be alarmed and can call for help immediately The system that will be created will lessen these huge damages at the same time will alert the owner immediately if the IP Internet Protocol camera saw an occurrence of fire or smoke There are a lot of studies regarding image processing it varies from article to article some are used for security purposes such as facial recognition for electronic camera devices and some are used in signal processing such as noise removal and filters In this study the researchers will be using image processing with covariance algorithms for fire detection systems A web camera will be used for capturing the video data needed for the system to interpret a fire and an IP camera that will serve as the receiving end of the system It will also allow the user to view the live video from the mobile
36. h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided 9 The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid 83 close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 Splot bo 1
37. ices do not work until the material comes in contact with the sensor and awareness of the additional information such as fire scale or location Today image processing has brought impact to the technology One of the common applications is facial recognition used on a laptop for security purposes Alert systems such as fire and other safety devices are possible to be combined with image processing technology Image processing usually refers to digital image processing Computer graphics and computer visions are very much related to image processing As people move forward to technology this study will provide data on the use of home alert systems as communication through mobile application and image processing technology The mobile application development for fire alert system allows the user s to be informed when a fire happens occurs inside their house There are fire detection systems that use smoke sensors which require a certain concentration of the smoke before it can detect fire but can lead to false detection such as a cigarette smoke The common problem where fire starts at home is caused primarily by homeowners being careless when it comes to dealing with potential hazardous situations Another cause usually deals with electricity where people forget to unplugged electric appliances The problem in the existing studies is accuracy in terms of time on when will the device detect the flame The system that will be created is very helpful
38. igVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen ex Red 300 Remove all those pixels less than 300px ex Green x Green medfilt2 ex Green 3 3 filter out noise o image into a binary image 9 less than 300px x Green bwareaopen ex Green 300 imsubtract greenVid rgb2gray videoFrame Use a median filter to ex Green im2bw ex Green 0 18 Convert the resulting grayscale Remove all those pixels ex Blue imsubtract blueVid rgb2gray videoFrame ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oo Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox MR mean mean redVid 60 Centroid MG mean mean greenVid MB mean mean blueVid S 0 3 Sa
39. igned to each sensor Most of the fire detection system used the heuristic fixed threshold values in their specific methods However input images may be changed in general so the heuristic fixed threshold values used in the fire detection systems might be modified on a case to case basis The image processing based fire detection has various advantages Light travels at very high speed so the image processing based fire detection can be quick to catch the spot of fire on time without delay Thus it can monitor wider space than the conventional sensor based fire detection system Most of the fire detection system used the heuristic fixed threshold values in their specific methods In this paper an automatic fire detection system without the heuristic fixed threshold values was studied It presented an automatic method using the statistical color model and the binary background mask Bo Ho Cho Jong Wook Bae Sung Hwan Jung 2008 12 A New Image Based Real Time Flame Detection Method Using Color Analysis This paper presents a new image based real time flame detection method The fire flame features based on the HIS color model are extracted by analyzing 70 flame images Then based on these flame features regions with fire like colors are roughly separated from an image Besides segmenting fire flame regions background objects with similar fire colors or caused by color shift resulted from the reflection of fire flames are also separated
40. ile application s notification 5 Repeat steps 2 4 for 20 times 6 Move the burning object to the next distance 7 Compute for the percentage of detection and percent error with the given formula 37 Number of Fire Detected NC RN 3 5 Percentage of Detection Total Number of Samples In calculating percentage of detection it is computed by the number of fire detected over total number of samples multiplied by 100 Percentage Error 10096 Percentage of Detection 3 6 The percentage error is calculated by subtracting percentage of detection from 100 Percentage of Detection E Percent Error Table 3 6 Graph of Distance Percentage Error 38 3 Android Mobile EE NEN E es aO A E O NN SN DENT SEN EN MEM MC NE Yes Yes Yes Yes Yes Table 3 7 Response of Android Mobile Application Table 3 7 shows the accuracy of the android mobile application as seen where the application detects the entire alert trial from trial 1 up to trial 10 But it differs in time it takes to alert the android mobile application The two major factors affecting the speed are the network provider and the speed of the mobile phone to receive SMS Even the time it will take to view the live streaming video will depend on these two major factors This means that the accuracy of the application doesn t rely on how the designers created the android mobile application Result of speed varies depending on those two factors
41. ire detection system which uses a spatio temporal co variance matrix of video data The system divides the video into spatio temporal blocks and computes co variance features extracted from these blocks to detect fire Feature vectors taking advantage of both the spatial and the temporal characteristics of flame colored regions are classified using an SVM classifier which is trained and tested using video data containing flames and flame colored objects Co variance method is ideally suited for flame detection because flames exhibit random behavior and it is observed that the underlying random process can be considered as a wide sense stationary process in a flame region in video The method uses temporal co variance information in the decision process Most fire detection methods use color spatial and temporal information separately but this method uses temporal extended co variance matrices representing all the information 14 together The method works well when the fire is clearly visible and in close range such that the flicker and irregular nature of flames are observable Fast and Efficient Method for Fire Detection Using Image Processing This paper presents a computer vision based fire detection algorithm In general a computer vision based fire detection system employs three major stages fire pixel classification moving object segmentation and analysis of the candidate regions This analysis is usually based on two figures the
42. ke There are a lot of studies regarding image processing it varies from article to article some are used for security purposes such as facial recognition for electronic camera devices and some are used in signal processing such as noise removal and filters In this study the researchers will be using image 20 processing with covariance algorithms for fire detection systems A web camera will be used for capturing the video data needed for the system to interpret a fire and an IP camera that will serve as the receiving end of the system It will also allow the user to view the live video from the mobile application This study aims to develop an algorithm for fire detection using shape and color analysis The specific objectives of this research are as follow a To create a mobile application that would alert and automatically inform the user if there is a fire detected b To develop fire detection system using image processing c To develop a more accurate algorithm that is early fire detection The study impacts on the safety of the people as well as contribute greatly in the safety field through the convenience of an alert system with the use of a mobile application The research and the proposed algorithm are suited to the advancement of technology where people nowadays are seeking portability of things The portability of the device will be evident in such a way that the user will be alerted from the mobile phone theoretically a
43. lete vid clear vid 2 MR MG break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG M text 0 1 MeanRGB end subplot 2 4 1 73 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 splot be 1 be 2 m a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end fwrite arduino angle3 Write to arduino angle 90 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red imsubtract redVid rgb2gray videoFrame ex Red medfilt2 ex Red 3 3 Use a median filter to filter out noise ex Red im2bw ex Red 0 18 Convert the resulting grayscale image into a binary image x Red bwareaopen
44. mework which enables easy creation of applications for different mobile devices and foster a GUI which reacts to context changes without having the need for any application logic supporting it It proposes an extension to XUL to enable the easy creation of applications which adapt themselves to different devices and user context The Java platform was chosen since it is available for most mobile phones and PDAs Thomas Butter Markus Aleksy Philipp Bostan and Martin Schader 2007 Based on the article Porting mobile web application engine to the Android platform Android is released as an open source mobile phone operating system is a Linux based platform it consists of the operating system middleware and user interface and application software xFace is a cross platform web application engine for mobile internet It uses HTML CSS and JavaScript and other Web technologies to develop lightweight mobile applications Mobile applications which are developed by using the technology of xFace can run on different operating systems without any modification such as windows Mobile BREW Symbian and other mobile phone platforms The application framework provides java class library for Android applications The libraries layer is developed by C C language where those libraries are packaged by Application Framework section The bottom section is the Linux Kernel which is developed by c language it provides the core system services such as se
45. mobile application s notification 5 Repeat steps 2 4 for 25 times 6 Change to another burning material 33 7 Compute for the percentage of correct detection and false detection with the given formula Number of Correct Detection Correct Detection x 10096 3 1 Total Number of Samples This formula is used in calculating systems correct detection Correct detection is calculated by the number of correctly detected samples over the number of total number of samples False Detection 10096 Correct Dtection 96 3 2 The number of false detection is computed by subtracting correct detection from 100 9 Correct Detection E False Detection Rubber Plastic Wood Table 3 2 Graph of Performance and Evaluations 34 ED I No of Faulty Detection System Performance Color Detection by RGB 19 104 18 26 81 74 color pattern Motion Detection by 17 104 16 34 83 66 comparing pixel Area Dispersion Detection 10 104 9 61 90 36 by checking deviation in Optimized Flame Detection 8 104 7 69 92 31 by images processing based techniques Proposed Fire Detection 5 104 4 81 95 19 System Table 3 3 Comparing previous results achieved in implementing different methods Table 3 3 presents the comparison of the proposed fire alert system versus the different methods used on different studies The number of faulty detection signifies the number of false detection o
46. nd SMS only but more usage of data to connect on the internet with lightning speed and much richer experience for users as they can go to social networking sites and online messaging with ease Nowadays so many mobile applications are created to monitor unexpected emergencies There were created because in today s generation almost everyone has their own mobile phone which can be easily accessed Android Applications According to the article Research on Development of Android Application android is a comprehensive operating environment based on Linux V2 6 kernel and is also a layered system Figure 2 1 Architecture of the Android System Application layer is the site of all Android applications including an email dient SMS program maps browser contacts and others Linux kernel is located at the bottom layer of the Android system and acts as an abstraction layer between the hardware and the rest of the software stack It provides core system services such as security memory management process management network stack and driver model A core feature of Android is that one application can use component element that belongs to another application In order to achieve such functions Android system must launch the application while any part of the application is asked and instantiated Java objects being asked Unlike most operating system there is no single point that the system can enter in an Android application Instead each
47. ndingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel TR subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROC Lr r 82 ESSING cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION 2f SR ST text 0 0 sprintf capsT Determine the camera deviceID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle
48. nywhere provided that the mobile phone is powered on The mobile application can be used instantly once the web server distinguish that the captured image of fire transmit a signal Lastly this research will contribute in the future studies of wireless technology which is now one of the most widely used technology in transferring 21 data or information The research also focuses on house monitoring of fire through image processing It alerts the user if there is a fire occurring in the said location The study covers the use of co variance based on fire color detection and shape analysis to determine a fire through image processing A web camera is used by the system to grab the video frames that are subject to be analyzed by the algorithm The system will capture a video clip once it detects a fire and stores it in its memory This will then send a signal to the android phone to automatically activate the mobile application for viewing and connect to the IP camera to display the activity in that location The mobile application must be installed in the mobile phone that comes within the system which is the only one capable of communicating to the IP camera The range of the camera is only limited to what it can see and the system cannot take action in stopping the unexpected emergencies 22 Methodology Hardware Development This section provides an overview of the steps that the designers undertook in the hardware development of thi
49. o clips to the Internet without using a PC so that people can remotely browse and watch the protected area activities Hence the performance of the IP surveillance cameras is generally evaluated at encoding speed video resolution video frame bitrates and distortion power dissipation etc Among these characteristics the video resolution the compression ratio and the frame rate are the most important to the users of the video surveillance systems because these features are highly related to the image quality and the network bandwidth MJPEG and MPEG4 Visual are the common standards used for image compression Ming Jiang Yang Jo Yew Tham Dajun Wu Kwong Huang Goh 2009 10 Live Video BF561 DSP ARM9 H 264 Streaming Encoder Network Interface Server Bit Streaming Figure 2 3 Block Diagram of IP Camera Related Studies Image Processing based Fire Detection System using Statistic Color Model Fire is necessary and profitable to human kind life but it also causes the damage to life and property in certain cases In general the early presence of fire can be detected by smoke gas or flame Most of the fire detection system equipment detects the fire based on sensors However a sensor cannot be reliable and fast because there can be times of delay with the sensor If the threshold value is low the sensor can have high sensitivity but it can produce false alarm And it also requires that a certain threshold value must be ass
50. one From the mobile application the user will be able to view the scenario in the area through the IP camera 18 Introduction Recently terrible disasters such as tsunami flood avalanche fire and explosion have caused many casualties and damages to property Therefore efficient detection for early alarming systems has to be used and has been an important research area in the last few years Researches on the automatic fire detection in particular has been widely studied Sensor based fire detection is being used due to its characteristics of low cost and simple set up However these devices do not work until the material comes in contact with the sensor and awareness of the additional information such as fire scale or location Today image processing has brought impact to the technology One of the common applications is facial recognition used on a laptop for security purposes Alert systems such as fire and other safety devices are possible to be combined with image processing technology Image processing usually refers to digital image processing Computer graphics and computer visions are very much related to image processing As people move forward to technology this study will provide data on the use of home alert systems as communication through mobile application and image processing technology The mobile application development for fire alert system allows the user s to be informed when a fire happens occurs inside their house
51. r to filter out noise 9 ex Green im2bw ex Green 0 18 Convert the resulting grayscale image into a binary image x Green bwareaopen ex Green 300 Remove all those pixels less than 300px ex Blue imsubtract blueVid rgb2gray videoFrame 9 ex Blue medfilt2 ex Blue 3 3 Use a median filter to filter out noise 9 ex Blue im2bw ex Blue 0 18 Convert the resulting grayscale image into a binary image o x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oe Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid d M mean mean redVid MG mean mean greenVid M S B mean mean blueVid 0 3 Saturation value of a pixel subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow 53 if MR gt MG amp amp MG gt MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING
52. ry vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k gt loopnumber delete vid clear vid break end 76 end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 f MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 Splot be 1 5Db9 2 Tmt a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts 1 else frameCounts 0 break end 77 fwrite arduino angle2 Write to arduino angle 45 degrees pause 0 02 videoFrame step videoDevReader origVid videoFrame redVid videoFrame 1 greenVid videoFrame 2 blueVid videoFrame 3 ex Red
53. s research Hardware development includes the flow of data through the components block diagrams and schematic diagram of the research as well as components used in the research The designers researched and gathered information about previous system as basis for circuit design They also researched components that can fit into the system to make it more cost effective without affecting the output From the information gathered the designers selected components to be used in the research Conceptual Framework Fire Detection System Fire Detection Algorithm Web Camera Mobile Phone Application Android Mobile Live Video Application IP Camera Streaming Figure 3 1 Conceptual Framework 23 Figure 3 1 shows the conceptual framework of the study For the fire detection system the web camera rotates within an area for image processing The fire detection algorithm analyzes the environment for possible fire occurrences and detects an object as a fire if it satisfies the algorithm The system will then save a short video clip for verification and send an alert message to the mobile phone through the internet When the mobile phone receives the message the mobile application will be automatically activated The mobile application will connect to the IP camera and present a live video streaming of the scenario in the mobile phone Block Diagram Figure 3 2 Block Diagram of the Moving Camera Figure 3 2 shows
54. shape of the region and the temporal changes of the region The fire detection performance depends critically on the effectiveness of the fire pixel classifier which generates seed areas than the rest of the system will exercise The fire detection algorithm consists of two main parts fire color modeling and motion detection The algorithm can be used in parallel with conventional fire detection systems to reduce false alarms It can also be deployed as a stand alone system to detect fire by using video frames acquired through a video acquisition device It is assumed that the image acquisition device produces its output in RGB format The algorithm consists of three main stages fire pixel detection using color information detecting moving pixels and analyzing dynamics of moving fire pixels in consecutive frames 15 Implementation of Object Tracking and Motion Sensors for Home and Office Based WIFI Monitoring System According to the design Implementation of Object Tracking and Motion Sensors for Home and Office Based WIFI Monitoring System the designers used an Arduino as a microcontroller for the IP camera s motion The microcontroller manages the data from the sensor and switches to properly move the camera The designers changed the medium of data transmission that makes the system portable and allowing for remote viewing anywhere with an internet connection The servo motor is the one responsible for generating the movement of the IP
55. t B Since fire is a light source its pixel values must be larger than some threshold R is the threshold for red channel R gt Rz Last is about saturation Sis the saturation value of a pixel and 5 is the saturation value of this pixel when R is RT S gt 255 R S R If the three conditions are satisfied then it will be classified as a fire Thus using co variance can include the location and intensity values of the pixels red green and blue values of the pixels as well as their first and second derivatives The system only covers a room environment set up Finally if it detects a fire the system will send a signal onto the mobile application for the user to be informed 30 T Yes Figure 3 6 Color Analysis Algorithm Figure 3 6 presents the color analysis algorithm used in the system The algorithm presents three conditions to be satisfied before it concludes that the algorithm has detected a fire A fire is an image that can be described by using its color properties This color pixel can be extracted into the individual elements as R G and B which can be used for color detection The first condition R z G B is computed from the means of red green and blue extracted from the video Where R should be more stressed than the other components because R becomes the 31 dominating color channel in an RGB images of flames R gt R is the second condition for R to be over some pre determined threshold R However
56. the block diagram for the hardware development of the system The moving camera is serially connected to the laptop and awaits the commands from the program and transmits the signal to the microcontroller The microcontroller controls the movement and timing of the servo motor to where it should focus at a certain time The web camera is mounted to the servo motor and serves as a tool for image processing 24 Review of Related Literature Integration of Hardware and Software Components Hardware Development Results and Discussion Development of Web Server Testing Development of the Fire Color Detection Algorithm Conclusion Development of the Mobile Application Figure 3 3 Procedural Steps in Conducting the Study Figure 3 3 shows about the procedures one must do in conducting the study First is researching about the related literature which will help by brainstorming for a possible and innovative topic The hardware development discusses the materials needed as well as the steps in designing the device The 25 development of the Fire Color Detection Algorithm and Mobile Application will fall under the category of software development In the software development the researchers determine what platform must be used and also the coding algorithms that will satisfy the software components of the system 26 Software Development The mobile application can be programmed using C C
57. tions takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memory vid DiskLogger aviobj vid TriggerRepeat Inf set vid FramesPerTrigger 1500 start vid close vid DiskLogger flushdata vid pause 6 clear my log refnum refnum 1 if k loopnumber delete vid clear vid break end end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R 2f G 2f B 2 MR MG MB text 0 1 MeanRGB end 80 else subplot 2 4 5 8 Visible off cla MeanRGB sprintf R text 0 1 MeanRGB oe oe 2 G oe 2 B 2 MR MG MB end subplot 2 4 1 for object 1 length stats bb stats object BoundingBox bc stats object Centroid rectangle Position bb EdgeColor r LineWidth 2 plot bc 1 bc 2 m a text bc 1 15 bc 2 strcat X num2str round bc 1 Y num2str round bc 2 set a FontName Arial FontWeight bold FontSize 12 Color yellow drawnow end if frameCounts 20 frameCounts frameCounts
58. tle Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION Sek SR OT text 0 0 sprintf capsT Determine the camera devicelID imagreset infom imaghwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end 79 filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided The following sec
59. tract Introduction Methodology Hardware Development Conceptual Framework Block Diagram Procedural Steps in Conducting the Study Software Development Shape and Color Analysis Algorithm Color Analysis Algoritm Testing and Statistical Treatment CHAPTER 4 CONCLUSION vi vi vii vii CHAPTER 5 RECOMMENDATION APPENDICES Appendix A Operation s Manual Appendix B Pictures of Prototype Appendix C Program Listing 44 45 46 49 52 LIST OF TABLES Table 3 1 Performance and Evaluations Table 3 2 Graph of Performance and Evaluations Table 3 3 Comparing previous results achieved in implementing Different methods Table 3 4 Graph of Comparing previous results achieved in implementing Different methods Table 3 5 Distance Percentage Error Table 3 6 Graph of Distance Percentage Error Table 3 7 Response of Mobile Application vi 33 34 35 36 37 38 39 LIST OF FIGURES Fig 2 1 Architecture of the Android System Fig 2 2 xFace Porting Layer Fig 2 3 Block Diagram of IP Camera Fig 2 4 Block Diagram Implementation of Object Tracking and Motion Sensor for Home and Office Based WIFI Monitoring System Fig 3 1 Conceptual Framework Fig 3 2 Block Diagram of Moving Camera Fig 3 3 Procedural Steps in Conducting the Study Fig 3 4 Android Mobile Application Algorithm Fig 3 5 Shape and Color Analysis Algorithm Fig 3 6 Color Analysis Algorithm vii 11 17 23 24 25 27 29 31 ABSTRA
60. tract blueVid medfilt2 ex Blue im2bw ex Blue 0 18 rgb2gray videoFrame 3 3 Use a median filter to Convert the resulting grayscale 67 9 x Blue bwareaopen ex Blue 300 Remove all those pixels less than 300px bw bwlabel ex Red 8 Label all the connected components in the image oe Here we do the image blob analysis We get a set of properties for each labeled region stats regionprops bw BoundingBox Centroid MR mean mean redVid MG mean mean greenVid MB mean mean blueVid S 0 3 Saturation value of a pixel subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvimage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION S128 SR ST text 0
61. turation value of a pixel A subplot 2 4 1 imshow origVid title Original drawnow subplot 2 4 2 imshow ex Red title Red Frame drawnow subplot 2 4 3 imshow ex Green title Green Frame drawnow subplot 2 4 4 imshow ex Blue title Blue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 9 Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 capsT sprintf RED SATURATION 2f nSATURATION Sek SR OT text 0 0 sprintf capsT Determine the camera devicelID imagreset infom imaghwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end 61 filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 Th
62. ue Frame drawnow if MR MG amp amp MG MB if MR 0 5 Threshold if S 1 MR 0 11 0 5 o Must examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvlImage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT 6 ST 1 MR 0 11 0 5 57 x capsT sprintf RED SATURATION 2f nSATURATION SLE SR ST text 0 0 sprintf capsT Determine the camera deviceID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj
63. ugh image processing A web camera is used by the system to grab the video frames that are subject to be analyzed by the algorithm The system will capture a video clip once it detects a fire and stores it in its memory This will then send a signal to the android phone to automatically activate the mobile application for viewing and connect to the IP camera to display the activity in that location The mobile application must be installed in the mobile phone that comes within the system which is the only one capable of communicating to the IP camera The range of the camera is only limited to what it can see and the system cannot take action in stopping the unexpected emergencies Chapter 2 REVIEW OF RELATED LITERATURE One of the most fast paced technology is mobile phones from Motorola Dynatac 8000x the first portable cellular phone which can be used to call and save 30 phone numbers to 2G GSM and CDMAone Cellular phones allow more calls to be packed into the same amount of radio bandwidth where one benefits to lower emission on the health an individual user The 2 4 generation networks provide easier communication that previous analog network on voice SMS and MMS Moving into the 3 and 4 generations from EDGE EVDO to HSPA Wi max and later on almost all wireless carriers are moving to LTE mobile phones improved be led by iPhone Android Phones and Windows Mobile With 3G and later on LTE users will not be restricted on using voice a
64. ust examine here for saturation subplot 2 4 5 8 Visible off capsT sprintf FIRE DETECTED PROCESSING cla text 0 1 capsT drawnow hsvimage rgb2hsv videoFrame Convert the imag to HSV space satRed hsvlimage 1 Get saturation red SR mean mean satRed Get mean S gt 1 R ST RT ST 1 MR 0 11 0 5 75 X capsT sprintf RED SATURATION 2f nSATURATION SLE SR ST text 0 0 sprintf capsT Determine the camera deviceID imaqreset infom imaqhwinfo for k 1 length infom InstalledAdaptors info imaqhwinfo infom InstalledAdaptors k if isempty info DeviceIDs Installedadaptor infom InstalledAdaptors k break end end if exist Installedadaptor var h warndlg Video camera not connected Warning replace pause 5 if ishandle h close h end beep return end filenameprefix FireAlertVideo Filename prefix adaptorname Installedadaptor deviceid 1 The maximum should be the total number of cameras connected loopnumber 0 How many time the video is divided The following sections takes video according to the time specified by the user k 1 while 1 my log filenameprefix num2str refnum avi aviobj VideoWriter my log Motion JPEG AVI vid videoinput adaptorname deviceid YUY2 320x240 vid LoggingMode disk amp memo
65. ver the total number of samples tested Number of Faulty Detection Percentage of Faulty Detection 100 3 3 Total Number of Samples In calculating percentage of faulty detection it is computed by the number of faulty detection over total number of samples multiplied by 10096 System Performance 10096 Percentage of Faulty Detection 3 4 35 The system performance is calculated by subtracting percentage of faulty detection from 100 Faulty Detection E System Performance Color Motion Area Optimized Proposed Detection Detection Dispersion Flame System Detection Table 3 4 Graph of Comparing previous results achieved in implementing different methods 36 Distance of the fire Percentage of Percent Error 9 o from the camera cm Detection 9 o 100 20 20 100 96 0 150 20 20 100 0 200 19 20 95 5 250 18 20 90 10 Table 3 5 Distance Percentage Error Table 3 5 presents the distance percentage error this testing measures the range of the system The system s accuracy is tested from different ranges in centimeter Each range is tested with 20 samples Procedures in conducting the distance percentage error testing are 1 Set up the fire alert system 2 Put the object to be burned in the said distance 3 Wait until the system detects the fire If the fire becomes bigger and the system doesn t detect this is considered it as a false detection 4 Wait for the android mob

Download Pdf Manuals

Related Search

Related Contents

F.スーツケース、アタッシュケース  Canon 2578A002 Camera Lens User Manual  TABLA DE CONTENIDO  半自動溶接機改造キット - 安川エンジニアリング株式会社  GSH 27 VC - Hizlial.com    Option Parts  Epson CX5200 Installation Booklet  ClosetMaid 12285 Instructions / Assembly  @7mudm - Makita  

Copyright © All rights reserved. Failed to retrieve file