Research On Video Detecting And Tracking For Pedestrian Counting Application

With the raising of public safe requirement, the video surveillance system gets the extensive application in many aspects in national life, such as safety supervises and controlling in intelligent communities and supermarket, the control of transportation discharge and so on. This paper mainly discusses the detection and tracking for counting of moving human in video sequences gathered from a fixed camera.On the research of the motion detection, the paper researched main algorithms of motion detection and analyzed advantages and drawbacks of these algorithms. This paper proposed an improved algorithm based on the principle of frames subtraction.On the research of object tracking, the paper researched several main object features and matching algorithms. Firstly, ordinary template-matching method without any improved methods is researched, and it is found that this method cannot be applied to tracking moving target at all due to huge time consumption. Therefore, Hausdorff matching is researched, the paper proposes a method of fast target matching based on Hausdorff distance, its main idea is that coarse matching is operated in the lower resolution image, obtaining some waiting regions, then the best matching target is attained by clusters from fast matching, and find the position of moving target. In the tracking process,target's velocity and direction measure of smoothness is used to create target-chain or tracking of moving target。Experimental results show that the method ensure the continuity and effectiveness of the dynamic pedestrians tracking and counting.Finally, the work of this dissertation is summarized and point out many existing problems about designing integrated targets track system frame. The paper makes some suggestions for further research on the realization of the system and the theories.