Design Of People Counting System On Intelligent Video Surveillance

In the field of intelligent monitoring, people counting problems is the current frontier topic which involve many disciplines and many complicated problem. Under the intelligent video surveillance on people counting can be used not only in shopping malls supermarkets business, it can also play an important role in terms of security, transportation, etc. The research of people counting system on intelligent video surveillance is to count people of the surveillance area with the technology of intelligent video surveillance, it includes the core technologies of motion target detection, human identification and tracking. For the requirement of open environment people counting, the author of this thesis design the people counting system based on intelligent video surveillance system, and research the core technology, the main work is as follows:(1)Through comparing with varieties of motion detection methods, the thesis selects the detection algorithm based on ViBe after evaluating and comparing the current mainstream. Aiming at the deficiency of traditional ViBe algorithm, an improved algorithm is proposed. Through to classify the foreground pixels and determine ghost shadow, the improved algorithm can eliminates the ghost and suppress the flicker noise pixels effectively.(2)In the aspect of the human body target recognition and tracking,take HOG features described as a pedestrian, combined with the improved ViBe algorithm to get the characteristic values of HOG in the may exercise area. And then use the trained SVM classifier to classify and recognize the human body. In the thesis, we use the Cam-shift algorithm to track human body. Aiming at target obscured when mismatch phenomenon, Cam-shift algorithm combined with Kalman filter to predict the target motion trajectory.(3)Based on the core technology above research, the thesis finally completes the design of the system hardware and software platform, and develops the client interface. Through the debugging and the experiment the system, the system can count people accurately, and it has stability very well.