The Research Of Algorithm Of People Counting Base On The Head Detection

Starting from the traditional infrared detection to the current intelligent video processingsystem，we count the number of people base on the detected target.but the Infrared detectionwill have shade effect seriously, specifically in the situation of a large number of people andpeople walking side by side. so some human body detection based algorithm can not count theperson’s number. Base on the problem above,we propose a new algorithm to solve thisproblem.First, we study some commonly used moving object detection methods: optical flowdetection, Interframe difference method, Background reduction method, three framedifference based of adaptive background update human detection algorithm, Head testing andson on. According to the characteristics of the various testing methods, we analyze theirapplicable places and their advantages and disadvantages respectively. Then choose the headdetection as number of statistical method. In terms of head count, we choose the Haar featurebased AdaBoost classifier detection method. In order to enhance the practicability of thesystem, we add the interested region setting, enhance the system operability. Finally, wedetermine a new staff to join in according to the distance clustering.If a distance between thecoordinate of a new staff and the coordinate of all detected people in previous frame is smallthan the threshold we set, then we consider it as a new staff join in.In order to verify the effectiveness of the proposed method, a large number of differentscenarios and different number of video are tested, results show that proposed algorithm isaccurate, real-time, effective, and has certain anti-interference ability.