Design And Implementation Of Pedestrian Detecting And Counting System In Channel Scenarios

With intelligent video surveillance develop rapidly nowadays,it has been applied to many aspects of our daily life. Intelligent video surveillance makes use of computer vision technology and image processing technology to detect object and try to understand their behaviors in image. And pedestrian detection and tracking is a hotspot in image processing research areas at present.This paper is about intelligent video surveillance and its technologies. We focus on sub-problem of pedestrian detecting and counting in Channel scenarios and made it into reality. We make researches and analysis on pedestrian detection, we developed a prototype system for pedestrian counting in channel scenerios combining with the existing research results. The main work of this paper includes the following five aspects:First,moving pedestrian detection algorithm,combined with fast three differential frames rapid object detection algorithm with head-shoulder contours,detects pedestrian quickly and meets the real-time requirements.Second,head-shoulder contour model detection method,we should make extraction of head-shoulder contour features at first,and matches the head-shoulder extracted out with samples in the head-shoulder contour model library,while matched,the target is detected out.Third, the human body tracking algorithm, making use of the head-shoulder contour detected out in the period of human body detection, tracks pedestrian.Combining head-shoulder features with the Kalman filtering, the algorithm can track pedestrian by tracking the head-shoulers part of human body and matching them with the head-shoulder sample library, it uses the minimum enclosing rectangle of human head-shoulder represents for the human body target.Fourth, multi-pedestrian tracking algorithm. While pedestrian in occlusion,overlapping,separation,the multi-pedestrian tracking algorithm can track pedestrian well.Fifth,the crowd counting algorithm counts pedestrian by extended of the double direction counting lines and achieve the total number of the crowd.The prototype system is developed on VS2008 platform in Windows operating system,using the OpenCV library and development language is C++,the hardware includes a camera, a laptop computer,of which the CPU frequency is 2.0GHZ and memory is 2G.