Research On Moving Object Detection And Tracking With Stereo Vision Technology In Static Scene

Stereo vision technology is a research hotspot in the field of computer vision.Among them,binocular stereo vision technology uses two cameras to simulate human eyes,and obtains two images of different angles,then obtains the specific 3D information of the target object in the actual space by the matching and other calculations.At present,this technology is widely used in mechanical workpiece detection,3D scene reconstruction,simulation AR,and other fields.According to the relative movement between the camera and the object,the scene can be divided into static scene and dynamic scene,the object detection and tracking in the static scene is researched in this article.The moving object detection and tracking algorithm based on monocular vision system is difficult to detect and track objects accurately because of shadow,occlusion and other problems.In order to solve these problems,binocular stereo vision technology is applied in moving object detection and tracking to overcome the problems of shadow and occlusion which cannot be overcome by single eye,and achieve real-time and accurate detection and tracking.The work of this paper is as follows:1)In camera calibration,this paper analyses the imaging model of the camera,including the linear and nonlinear model.Then,several common camera calibration methods are analyzed.The Zhang Zhengyou plane calibration is mainly studied in,and the relatively accurate camera's internal and outside parameters are acquired through experiments.2)In the binocular stereo matching,the common methods of the stereo matching are analyzed in this paper.According to the influence factors,the paper mainly study the SIFT matching algorithm.Due to the matching speed and accuracy,this paper adopts an improved algorithm based on SIFT algorithm.PCA algorithm is introduced to reduce the amount of computation,and RANSAC algorithm is introduced to reduce the matching errors.3)In moving object detection,an improved detection method is introduced after comparing the common traditional detection methods.The disparity map by binocular stereo matching is applied to the background differencemethod to obtain the background difference method based on the disparity map.Then the background difference method based on the disparity map combined with the background difference method based on the gray map accurately detect moving objects.4)In moving object tracking,the paper mainly study the MeanShift tracking algorithm.Because the algorithm is prone to the tracking failure in the case of occlusion,the paper combine binocular stereo vision technology and MeanShift tracking algorithm,then joined the Calman filter to predict motion state of the moving object.The method can predict the movement status of moving objects,overcome the occlusion problem,and realize the real-time accurate tracking.