Design And Realization Of Passenger Flow Analysis System Based On People Detection

With the development of video monitoring system, intelligent process is become more and more important. Passenger flow is indispensable data for management and decision in public places. Based on video monitoring system, process the video frame and count people number intelligently, has high research values and broad market prospects.At present, people counting system requires the camera set vertical above, counting the passenger by the combine of people tracking and counting line. But this method has strict requirement of camera and passenger. This paper proposes a novel system for passenger flow analysis, it processes the video monitoring image of entrances of public places. Through the real time processing of video image get from monitor devices, detecting the faces and combining with setted people counting area to realize automatic people detection and face area-tracking. The system can statistic the passenger flow accurately and analysis the result.The main contents of this paper are as follows:1. Aim at solving the problem that Adaboost algorithm do not perform well when the background is complicated, the paper utilize skin detection to improve the detection result. Though skin detection method can greatly reduce the false detection rate and improve the accuracy rate.2. In order to solve the problem that appeared in traditional people counting system, face area-tracking is proposed in passenger flow analysis system. Compared with the traditional method, setting the tracking area can reduce the computation burden and realize the real-time processing. And elements such as the position of the camera, the direction and path of passenger is not be strictly limited.3. So as to improve the business value of the system, the counting result is well processed. The result will be displayed in the interface directly as well as be stored in the sever. Beside, the counting-result query function make the system can be conveniently used in public places, through the analysis in the client, the statistical information will be displayed in the interface, so that the result is much more intuitive and effective.4. Test the performance of the system. The performance of the system is judged by the false rate, miss rate and accuracy rate. Through the test in the laboratory, experimental results show that this system has good real-time performance and high accuracy.