Study Of People-counting System Based On The Recognition Of Quasi-circle Graphics

A customer statistics method based on quasi-circle pattern recognition has beenproposed in this paper, according to the characteristics of the image of people’s top, inorder to meet the statistical analysis of customer information demand of the public place.The statistical method can solve moving target the identification and tracking problemsof the video surveillance area under complex background. The proposed method alsohas very important theoretical and practical significance for the peer malls, libraries,tourist attractions and other public places fast traffic statistics.The specific target identification and tracking of the matching count in thecontinuous video images is part of important research in pattern recognition and digitalimage processing, which includes three parts, namely as moving target detection, targetclassification and identification, target matching and counting.Firstly, there may be traffic statistics system based on image acquisition targetblock, repetition, shadow change and other issues caused by the light. To solve theproblem above, the camera is mounted above the import and export of shopping malls,and we can adjust the bracket so that the camera and perpendicular to the ground to gethuman head vertical target avatar and filter out some of the interference, which laid thefoundation for subsequent identification and tracking targets.Secondly, according to the quasi-circle imaging characteristics of the top image ofpeople, taking into account the color of his clothes, wearing a hat, carrying the impactitems and illumination change, a quasi-circle target detection algorithm based on theupdated background subtraction division is proposed by studying and comparing someof the algorithms used in target detection. The algorithm combines geometriccharacteristics of the target to be detected, improving the practical application of therecognition speed and accuracy.Finally, a tracking and matching algorithm based on quasi-circle targets isproposed, considering the actual movement of the camera tracking pedestrian area, witheigenvalues quasi-circle targets. The velocity estimation algorithm and matching costfunction reduce the traditional target tracking and matching algorithm complexity in thecase of ensuring the accuracy of traffic statistics. According to the quasi-circle target detection and tracking matching algorithm, wehave designed the traffic statistical systems under Windows platform, which is based onVisual Studio2010and OpenCV. After a lot testing of the captured video images, thedetection accuracy results come to be satisfactory. The statistical results are basicallyconsistent with the fact, which suggests that the system is feasible in practicalapplications.