Target Tracking Algorithm And Its Application In The District’s Video Surveillance

video monitoring technology which is treated as the important part of security system is paid more and more attention to. Accurate multiple motion targets detection and tracking are the foundation about people’s analysis and comprehension for suspicious actions, which has an important influence on the intelligent video surveillance. The difficulties in the tracking algorithm for targets are the problem of exact location for targets, the problem of shading and real time tracking of multiple motion targets. According to these difficulties, it designs a relative tracking algorithm for targets and estimates the performance of the algorithms. And, It studies on the application of these algorithms in the intelligent video surveillance in plot.For the shortage about color histogram dealing with targets shading problems of multiple motion targets tracking, it designs a kind of modified Adaptive appearance model for describing motion targets. The model includes color model and corresponding probability model. Because it includes target’s color information but also target’s spatial information, it could solve target location, correspondence and shading in tracking procedure.The particle filter algorithm for the multiple targets tracking needs lots of particles to estimate the condition of system,which lead to complicated calculation and bad real-time performance. Therefore, it designs a multiple targets tracking algorithm based on adaptive appearance model under the Bayesian theory frame. After targets detection, it could gain the mass movement. Moving targets is related by Mass data matching. And, solution of target shading is detected. Under the shading, particle filtering is created for every shaded target to track. The alrigothm is tested by PETS2001video sequence on the laptop whose frequency of KuRui dual-core is1.73G and dealing speed is about20-25frames per second. Of course, the performance of tracking is perfact. It gives performance evaluation framework about tracking algorithm. And, it designs evaluation indicators for tracking algorithm. At last, Algorithm performance evaluation and comparison are carried on for the tracking algorithm by experiment, which is designed by ourselves. The method could solve difficulties of manual test. That people’s subjective feeling and emotional factors have influence on evaluation results makes the repeatability of evaluation results much worse.For real scene of the intelligent video surveillance in plot, multiple target tracking algorithm and real video surveillance system are connected. A kind of prototype of intelligent video surveillance system is developed by ourselves. Intrusion detection and loitering detection in the warning region are treated as examples. It studies implementation ways and application scene about function of system, whose server-side is developed on the VS2010and the man-machine interface is friendly which is so convenient for users to set the warning region and check the alarm log. it provides theories and engineering application plat for the algorithm.