Research On Moving Human Tracking And Behavior Analysis In Video Surveillance

With the development of computer technology, and also intelligentization of video surveillance, computer vision-based intelligent video surveillance system catches more and more attention among domestic and foreign researchers. The intelligent surveillance system can not only automatically record the scene of incidents, but also raise the alarm immediately when abnormal behavior and events occur. This will improve the efficiency of the surveillance system and save a lot of human and material resources, as well as bring our society with tremendous economic benefits. Vision based motion analysis of the human body is a key issue in this research area, such as intelligent video surveillance, airport abnormal event monitoring and other areas, which has a wide application.This paper mainly studies the movement of video surveillance system analysis and event detection of human behavior. Firstly we apply three consecutive frame difference method and adaptive background updating algorithm to detect moving human body, and then process morphologically the detected moving human to eliminate noise and small empty, and also detect and eliminate shadows. The next step is to track the detected moving human body on the basis of the analysis of the current object tracking method. Since the human belongs to non-rigid moving object, we propose a hybrid algorithm of mean shift and particle filter tracking. The final step is to recognize the behavior of moving human body. We propose an improved Hu invariant moments of abnormal behavior recognition algorithm, and make a brief research on specific events at a particular environment. The experimental results show the proposed algorithms are robust and possess high performance.