Video Surveillance Systems For Traffic Monitoring


Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Video Surveillance Systems For Traffic Monitoring

  1. 1. Video Surveillance systems for Traffic Monitoring Simeon Indupalli
  2. 2. Presentation Overview <ul><li>Video surveillance systems. </li></ul><ul><li>Traffic monitoring issues. </li></ul><ul><li>Object tracking techniques. </li></ul><ul><li>Vehicle tracking strategies. </li></ul><ul><li>A real time system Explanation. </li></ul><ul><li>Future Work </li></ul>
  3. 3. what is video surveillance? <ul><li>Present Implementations? </li></ul><ul><ul><li>Human detection systems. </li></ul></ul><ul><ul><li>vehicle monitoring systems. </li></ul></ul><ul><li>Advantages of video surveillance? </li></ul><ul><ul><li>Keep track of information video data for future use. </li></ul></ul><ul><ul><li>Helpful in identifying people in the crime scenes etc.. </li></ul></ul><ul><li>Disadvantages of the present system? </li></ul><ul><ul><li>It’s difficult to maintain heavy amount of raw video data </li></ul></ul><ul><ul><li>Human interaction. </li></ul></ul><ul><ul><li>Require higher bandwidth for transmitting the visual data. </li></ul></ul>
  4. 4. Video surveillance in the context of Computer Vision <ul><li>Detection and tracking of moving objects are the important tasks of the computer vision. </li></ul><ul><li>The video surveillance systems not only need to track the moving objects but also interpret their patterns of behaviours. This means solving the information and integration the pattern. </li></ul><ul><li>Advantages </li></ul><ul><ul><li>Minimizes the user interaction. </li></ul></ul><ul><ul><li>Less amount of prohibitive bandwidth. </li></ul></ul><ul><ul><li>Minimizes the cost and time. </li></ul></ul>
  5. 5. Need for Traffic Monitoring <ul><li>To reduce the traffic congestion on highways </li></ul><ul><li>Reduce the road accidents </li></ul><ul><li>Identifying suspicious vehicles. Etc.., </li></ul>
  6. 6. Traffic Monitoring in Computer Vision <ul><li>The quest for better traffic information, an increasing reliance on traffic surveillance has resulted in a better vehicle detection. </li></ul><ul><li>Taking some intelligent actions based on the conditions. </li></ul><ul><li>Traffic scene analysis in 3 categories. </li></ul><ul><ul><li>A strait forward vehicle detection and counting system . </li></ul></ul><ul><ul><li>Congestion monitoring and traffic scene analysis. </li></ul></ul><ul><ul><li>Vehicle classification and tracking systems which involve much more detailed scene traffic analysis. </li></ul></ul>
  7. 7. Responsibilities of reliable Traffic Monitoring System <ul><li>Adaptive to changes in the real world environments </li></ul><ul><li>Easy to set up </li></ul><ul><li>Capable of operating independently of human operators. </li></ul><ul><li>Capable of intelligent decisions. </li></ul><ul><li>Capable of monitoring multiple cameras and continuous operation. </li></ul><ul><li>Reasons for unsuccessful implementation** </li></ul>
  8. 8. A Traffic Monitoring System
  9. 9. Object Classification <ul><li>Shape based classification. </li></ul><ul><ul><li>Image blob area, blob bounding box </li></ul></ul><ul><ul><li>Classification based on above info. </li></ul></ul><ul><li>Motion-based classification. </li></ul><ul><ul><li>Human motion shows periodic property. </li></ul></ul><ul><ul><li>Time frequency analysis applied. </li></ul></ul><ul><ul><li>Residual flow taken under consideration. </li></ul></ul>
  10. 10. Object tracking strategies (I)* <ul><li>Background subtraction </li></ul><ul><ul><li>Difference between the current image and the reference background image in a pixel by pixel fashion. </li></ul></ul><ul><ul><li>Sensitive to the background changes </li></ul></ul><ul><ul><li>Wallflower principles for effective background maintenance. </li></ul></ul>
  11. 11. Object tracking strategies (II) <ul><li>Temporal differencing </li></ul><ul><ul><li>Moving objects changes intensity faster than static ones </li></ul></ul><ul><ul><li>Uses consecutive frames to identify the difference. </li></ul></ul><ul><ul><li>Adaptive to dynamic scene changes </li></ul></ul><ul><ul><li>Problems in extracting all relevant features. </li></ul></ul><ul><ul><li>Improved versions uses three frames instead of two </li></ul></ul>
  12. 12. Object tracking strategies (III) <ul><li>Optical flow </li></ul><ul><ul><li>To identify characteristics of flow vectors of moving objects over time. </li></ul></ul><ul><ul><li>It’s used to detect independently moving objects in presence of camera. </li></ul></ul><ul><ul><li>Requires a specialized hardware to implement. </li></ul></ul>Optical flow of moving objects Meyer et al
  13. 13. Vehicle detection techniques <ul><li>Model based detection </li></ul><ul><li>Region based detection </li></ul><ul><li>Active contour based detection </li></ul><ul><li>Feature based detection </li></ul>
  14. 14. Vehicle detection technique (I) <ul><li>Model based Tracking </li></ul><ul><ul><li>The emphasis is on recovering trajectories and models with high accuracy for a small number of vehicles. </li></ul></ul><ul><ul><li>The most serious weakness of this approach is the reliance on detailed geometric object models. </li></ul></ul><ul><ul><li>Disadvantage </li></ul></ul><ul><ul><li>It is unrealistic to expect detailed models for all vehicles that could be found on the roadway </li></ul></ul>
  15. 15. Vehicle detection technique (II) <ul><li>Region based tracking </li></ul><ul><ul><li>It detects each vehicle blob using a cross correlation function. </li></ul></ul><ul><ul><li>Vehicle detection based on back ground subtraction. </li></ul></ul><ul><ul><li>Disadvantage </li></ul></ul><ul><ul><li>Difficult to detect the vehicles under congested traffic, because vehicles partly occlude with one another </li></ul></ul>Potential segmentation problem
  16. 16. Vehicle detection technique (III) <ul><li>Active contour based detection </li></ul><ul><ul><li>Tracking is based on active contour models, or snakes. </li></ul></ul><ul><ul><li>Representing object in bounding contour and keep updating it dynamically. </li></ul></ul><ul><ul><li>It reduced computational complexity compared to the region based detection. </li></ul></ul><ul><ul><li>Disadvantage: </li></ul></ul><ul><ul><li>The inability to segment vehicles that are partially occluded remains a problem. </li></ul></ul>Bounding counters
  17. 17. Vehicle detection technique (IV) <ul><li>Feature based detection </li></ul><ul><ul><li>Tracks sub-features such as distinguishable points or lines on the object </li></ul></ul><ul><ul><li>Effectiveness improved by the addition of common motion constraint. </li></ul></ul>Features are grouped together based on common motion, avoiding segmentation problem due to occlusion
  18. 18. A typical vehicle tracking procedure
  19. 19. Wallflower Principles & Practice of Background Maintenance. <ul><li>Moved objects </li></ul><ul><li>Time of day </li></ul><ul><li>Light switch </li></ul><ul><li>Waving trees </li></ul><ul><li>camouflage </li></ul><ul><li>Foreground capture </li></ul><ul><li>Stopped car </li></ul><ul><li>Moving car </li></ul><ul><li>Shadows </li></ul><ul><li>Bootstrapping </li></ul>
  20. 20. Wallflower: Three levels of abstraction <ul><ul><li>Pixel level </li></ul></ul><ul><ul><ul><li>Maintains models of back ground of each individual pixel. </li></ul></ul></ul><ul><ul><ul><li>Processing makes the preliminary classification between foreground and background </li></ul></ul></ul><ul><ul><ul><li>Dynamic to scene changes. </li></ul></ul></ul><ul><ul><li>Region level </li></ul></ul><ul><ul><ul><li>Emphasis is on interrelationship between the pixels </li></ul></ul></ul><ul><ul><ul><li>Helps to refine raw classification at pixel level </li></ul></ul></ul><ul><ul><li>Frame level </li></ul></ul><ul><ul><ul><li>It watches for the sudden changes in the large parts of the image and swaps in alternative background models. </li></ul></ul></ul>
  21. 21. A real time traffic monitoring system Feature based tracking algorithm <ul><li>Camera calibration </li></ul><ul><li>Feature detection </li></ul><ul><li>Vehicle tracking </li></ul><ul><li>Feature grouping </li></ul>Benjamin Coifman, Jitendra Malik, David Beymer
  22. 22. Offline camera definition <ul><li>Line correspondences for a projective mapping. </li></ul><ul><li>A detection region near the image bottom and an exit region at the image top </li></ul><ul><li>And multiple fiducial points for camera calibration </li></ul><ul><li>Based on the above information the system computes the homography between the image coordinates(x,y) and the world coordinates(X,Y) </li></ul>
  23. 23. On-line tracking and grouping <ul><li>Detector </li></ul><ul><ul><li>Detecting corners at the bottom of image, where brightness varies in more than one direction. </li></ul></ul><ul><ul><li>Detection operationalzed by the points in the image I </li></ul></ul><ul><li>Tracker </li></ul><ul><ul><li>Uses kalman filters to predict the velocity in the next image. </li></ul></ul><ul><ul><li>Normalized correlation is used to search the small region of image. </li></ul></ul><ul><li>Group </li></ul><ul><ul><li>Grouper uses common motion constraint. </li></ul></ul><ul><ul><li>Once all the corner features are identified they are grouped together. </li></ul></ul><ul><ul><li>Monitoring the distance between the point d(t)=P1(t)-p2(t) </li></ul></ul>
  24. 24. Sample corner features identified by the tracker Sample feature tracks from the tracker Sample feature groups from the tracker 1 2 3
  25. 25. Conclusion & Future Work <ul><li>The real time traffic surveillance system is still under research due to the background maintenance problem and occlusion. </li></ul><ul><li>Better Background maintenance </li></ul><ul><li>Solving occlusion problem </li></ul>
  26. 26. References: <ul><li>A Survey on visual surveillance of object motion and behaviour </li></ul><ul><li>– HU et al </li></ul><ul><li>Transportation research part-c/ A real time computer vision system for Traffic monitoring and vehicle tracking – B.coifman, J.Malik etc.. </li></ul><ul><li>Steps towards cognitive vision system – H.Nagel, IAKS Karlsruhe. </li></ul><ul><li>VSAM project – C arneigh Mellon University </li></ul><ul><li>Wallflower Principles and practices – Microsoft Research group. </li></ul>