You are on page 1of 4

International Journal of Research and Development Vol I ,Issue III, ISSN - 2279-0438

Computer Vision
Surya Mani Sharma Bachelor of Technology (Information Technology) Mahrishi Dayanand University Dronacharya College of Engineering Vision_suraj@yahoo.com

Abstract: The research work on this paper explains the robotic vision used in my MARS (Multi functional Autonomous Robotic System ) project[1,6] for visual recognition , tracking the objects and keeping eye on moving object and to develop computer memory for future use . Computer vision is an immense field which deals with every field of engineering vision system and video processes. Vision algorithm demonstrated in the paper are easier to implement and takes low amount of memory and process using the backend OpenCV library[2] which makes those technique efficient and reliable for low power system , mobile system and low cost system . Most of the filters used the technique are inherited from the OpenCV library. With months of research and study of various other vision algorithms and project it was possible to develop the below given technique. With the help of these techniques we can use our system to keep an eye on any desired object; it can also be used for actuator control solely based on camera. Way to implement the technique in binocular vision system [3] is still in raw phase. Those technique demonstrated is solely based on monocular vision so, for depth perception other sensor and mathematical calculations is used rather than image processing. 1 Introduction Computer vision is concerned with the automatic extraction, analysis and understanding of useful information from a

single image or a sequence of images. It involves the development of a theoretical and algorithmic basis to achieve automatic visual understanding. The applications of computer vision are numerous and include: augmented reality , autonomous vehicles , biometrics , character recognition , forensics , industrial quality inspection , face recognition , gesture analysis , geosciences , image restoration , medical image analysis , pollution monitoring , process control , remote sensing , robotics Security and surveillance etc. Whereas the demonstration in this paper is solely based on the application robotics and autonomous vehicle system [4] which can be used in various field as illustrated above. 1.1 Motivation There have been tremendous work in the field of computer vision and as eyes are the most essential part of the human body so is the vision for embedded system robots and units .Today computer vision is said to work as third eye of human beings either that is in industry, military, surveillance or research program. The field computer vision and processing is filled with energy and thrills in research sector. The most advanced application is seen in War in terror in Afghanistan where the autonomous UAVs are replacing the human soldier day by day [5]. Application of Computer vision system to empower the physically challenged people has also remained the main priority in research whose market is thought to be immense and awarding.

International Journal of Research and Development Vol I ,Issue III, ISSN - 2279-0438
2. Object Tracking Object tracking is one of the prime goals of the computer vision system; to follow up the object in the view provides the human like capability to the machine system. This paper investigates the way to trace object with respect to the location of that object in the frame according to its center of gravity with respect to X and Y coordinates for tracking in XY axis. Fig 1.1 shows the MARS system base robot which uses the same concept as mentioned to trace the ground objects or vehicle that are the prime target of the system and the code of it in VBScript is given below. end if end if else if cogx > 80 and cogx < 240 and coga > 70 and coga < 150 then led2 = 0 , led3 = 0 , led4 = 0 , led5 = 0 , end if end if if cogy > 150 or cogy < 70 then if cogy < 70 then led7 = 1 else if cogy > 150 then led6 = 1 end if end if else if cogy < 150 and cogy > 70 then led7 = 0 led6 = 0 end if end if In order to find the center of gravity of the system camera video feed is filtered with RGB filter with minimum desired size defined .After filtering the feed variables are used in the VBScript to control the movement . 3. Object Recognition Object recognition is said to be early AI capability of the machine system . With the OpenCV backend , after Haar training ; we can use the system video feed for detecting the objects appearing in the camera frame . Although the below given Python based Object recognition doesnt use the haar method for detecting the object from the still image which makes its efficient in static scenario application . Object recognition is most memory and Processing hungry task of computer vision system . Even then with availability of modern fast system with more memory its still a field growing in rapid phase . One of the recent module by name AVR navigator incorporates very efficient frame filtering mechanism and object learning system which is very good for low power system .

Fig 1.1 MARS Base Unit


VBScript code for object tracking in XY axis

if cogx < 80 or cogx > 240 or coga < 70 or coga > 150 then if cogx < 80 then led2 = 1 , led5 = 1 else if cogx > 240 then led3 = 1 , led4 = 1 end if end if if coga < 70 then led3=1 , led5=1 else if coga > 150 then led2=1 , led4=1

International Journal of Research and Development Vol I ,Issue III, ISSN - 2279-0438
Python code for image recognition from still image

a=cv.LoadImage("/home/jayneil/beautifulfaces.jpg") print(a) cv.NamedWindow("Original Image", cv.CV_WINDOW_AUTOSIZE) cv.ShowImage("Original Image", a) cv.WaitKey(6000) cv.DestroyWindow("Original Image") hc = cv.Load("/home/jayneil/haarcascade_frontalfac e_default.xml") img = cv.LoadImage("/home/jayneil/beautifulfaces.jpg", 0) faces = cv.HaarDetectObjects(img, hc, cv.CreateMemStorage()) for (x, y, w, h), n in faces: cv.Rectangle(img, (x, y), (x+w, y+h), 255) cv.SaveImage("faces_detected.jpg", img) print(faces) dst=cv.LoadImage("faces_detected.jpg") cv.NamedWindow("Face Detected", cv.CV_WINDOW_AUTOSIZE) cv.ShowImage("Face Detected", dst) cv.WaitKey(5000) cv.DestroyWindow("Face Detected") //Graphics Libraray : OpenCV // Python 3.Field Test Computer vision in my context developed while working for MARS (Multifunctional Autonomous Robotic System ) project in premise of Dronacharya College of Engineering , India . Base robot of MARS equipped with computer vision object recognition and tracking was aimed for tracking the desired ground object as well as recognizing the environment around it . The first test was carried in the Lavatory in which the robot task was to follow the flower moving in floor in xy plane , the robot with object tracking module did successfully tracked the object as desired but when the object would be too far or too near it encountered the

error which was avoided by using the another parameter from the filter COG_AREA , after that it the test was successful .

Fig 1.2 Image Recognition in process As the vary day different objects like mobile , multimeter , flower and other object were scattered in the floor and the task of the robot was to recognize them as it passes through them , robot was successful in achieving the desired result . So after the multiple testing and changes in system the base robot was able to achieve the real world viability . Due to few factors like Camera resolution , zoom capability and actuator power factor the system suffered many set backs . Camera resolution is an essential factor while working in low light area although the high resolution means more processing capability requirement but the camera of 10-15 megapixel is quite favorable for object recognition in low light imagery . MARS project has been a quite success which has opened new dimensions for the betterment of computer vision system in coming future .

International Journal of Research and Development Vol I ,Issue III, ISSN - 2279-0438
4. Conclusion and Further works : [3] books.google.co.in/books?id=I8vqITdETe0 C&pg=PA75&lpg=PA75&dq=in+binocular+vision +system&source=bl&ots=JghJY0FqNt&sig=dfHH MrDcN8n35iixZTzsFtZriUI&hl=en&sa=X&ei=LtfN UYfKL4nVrQei7IDIDA&redir_esc=y [4] http://idealab.talkingpointsmemo.com/ 2013/01/toyota-test-driving-autonomousvehicles-in-michigan.php [5] http://www.armytimes.com/article/ 20101216/NEWS/12160320/ArmyTremendous-demand-UAVs-Afghanistan

In this paper a brief overview about the computer vision system was presented , as well as brief overview of codes . Computer vision system finds its position more and more concrete with advancing technological implementation in all sectors of our life . Either it is surveillance or battle zone or any scientific survey of location and remote surveying .

The above given techniques of computer vision in adjacent with the openCV library is very sturdy to use in our day to day embedded system which can further be refined for according to need of time and workspace . Although the main part of the computer vision has remained the camera used which is essential to make sure the frames that are received in processing unit are as it exists in real world as we can see in the UAV systems . Computer vision with autonomous learning can be a third eye for human and will be a boon for industry as well as physically impaired people .

5. References :-

[1] ijisme.org/attachments/File/v1i6/ F0304051613.pdf [6] www.csiindia.org/c/document_ library/get_file?uuid=a6cf8d80-1065-456ab483-b6f21c73b636&groupId=10157 [2] opencv.willowgarage.com/wiki/

You might also like