Particle Filter Based Object Tracking System For Intelligence Video Surveillance

Intelligence Video Surveillance (IVS) using computer vision method to actualize the analysis of video content. It can identify special object or action and alarm when it found suspect object to actualize preventive surveillance. IVS technique has important applied value both in military and civil affairs. Object tracking is a crucial component of monitoring system. The tracking result affect the behavioral analysis witch is the based of IVS application. The study on object tracking technique is meaningful.Difficult tracking problems includes object non-rigid structure, object anomalous movement, object characteristic change and mutual mask of objects or shelter between object and obstacle etc. This paper studies on the object tracking under complex scene, for the purpose of proposing a tracking method that is effective and has good accuracy and robust. This article takes the object tracking as the research background and particle filter (PF) as the study object, realizes the track algorithm base on PF and carries on the algorithm optimization. The prime task includes:1) The study and implementation of the tracking algorithm based on PF.Do research on correlation theories of PF, than use them on video image to realize tracking algorithm, describe the algorithm flow in detail, establish movement model and solve problems particle update as well as particle re-sampling.2) Algorithm optimization.In order to solve difficult tracking problems as well as raise the algorithm efficiency, this article carries on various improvements on the algorithm. Mainly in the below aspects:(1) Adopt integral image method to speed up the histogram operating and then raise the algorithm efficiency;(2) Make use of foreground in doing object tracking based on PF, solve the problems background disturbing and shelter between object and obstacle;(3) Carry on the partition characteristic histogram statistics to the object. An object can be represented by color histogram with upper part and lower part. Make brightness compensation to the histogram so that the color histogram can better express object. In solution the problem color histogram change as the environmental variation this method has manifested its superiority;(4) Do shelter judgment of objects to solve the object mutual mask problem. The experiment proved that the optimized tracking algorithm based on PF has solved well in the difficult tracking problems such as background disturbance, illumination change and shelter.