Moving Object Tracking And Abnormal Behavior Recognition In Real-time Video Surveillance

Intelligent Video Monitoring is a technology which uses automated video analysis technology to carry out video monitoring. As an effective means of security guard, it takes more and more attention.Aiming at the application requirements of Intelligent Security Monitoring System, this paper designed a relevant human motion analysis system, which succeeded in detecting and tracking moving human body in real environment. Base on this, this system finished recognizing some exceptional human activities. By doing all these work, the Security Monitoring System become really "intellectualized".On the research of detection and extraction of moving objects, firstly, this paper summarized several algorithms currently, and juded the merits and faults of every algorithm. Based on these, this paper supposed an improved algorithm. Combining the mixed Guassian background modeling with Temporal difference, we could get more precise moving-human contour. Further more, we wiped off the shadow and the chirp from the foreground, via selective update, the "hole" problem caused by this algorithm was solved.On the research of human behavior recognition, this paper used two different methods. One is the traditional technology of Template Matching, which regards the human contour as a basic action and builds up a human activity sequence template. Using snake active contour line model, this system extracted accurate contour, and avoided the internal discontinuity effectively. Complete image of the moving target can be gained by filling this contour with white pixels. Comparing these images with the template, we can get simple human activity recognition. Another one is an algorithm which combines the center of mass with rectangle aspect ratio. Comparing with the technology of Template Matching, this method is simpler and more effective, and has higher test accuracy.