People Counting System Based On Video Surveillance

Aiming at the urgent demand of people flow counting from all over the world, this paper is devoted to develop a new people flow counting system which based on video surveillance. As an important application and research hotspot field of video surveillance, the people flow counting system was widely used in the stations’security, shopping malls, supermarkets and traffic management etc.. The system is mainly based on computer vision and pattern recognition technology, and finally realized the people flow counting by detailed research about detection algorithms for moving target and recognition-tracking algorithms for human object.(1) Through comparing with varieties motion detection methods, the paper selects the three frame difference method which belongs to the frame difference method, to be used as the core algorithm of the target detection. And we also combine the maximum between class variance method to adaptive the fixed threshold. Meanwhile, in order to reduce the time complexity of the improved algorithm, we increases a determine operation before the threshold change, so that the system could full improve the algorithm’s real-time and ensure the detection effect.(2) In the human target recognition, system adopts the method of SVM to classify the HOG features. Through extracting the HOG gradient information and SVM trains the samples, we could determine the detected targets whether to belong to human body.(3) For the target tracking part, the paper carefully analyses the defects of the commonly used Cam-Shift tracking algorithm, and puts forward the corresponding improvement. For the problem of complex background, the system adds multi-feature fusion to full extract the targets’ information. For the problem of occlusion interference, the system adds the improved Kalman filter to predict the potential motion trend.(4) Based on the research of the above key technology, the paper successful completes the building of the system’s hardware, software and the client platform. Finally, the practical test shows that the system can real-time and reliably counts the people flow, and has the good practical value.