Research On Moving Object Recognition And Tracking Based On Binocular Vision

Based on binocular vision,the moving target tracking system is based on binocular vision technology,and the process of moving objects in the video image sequence is detected and extracted,and the process of continuous tracking is carried out.The main process is to obtain the image with the depth information through the binocular camera,and combine the depth image and the grayscale image to detect the moving object to calculate the position and size of the moving object in each frame image.Then,the same characteristics of the target associated with the target location,speed,acceleration and other information.Finally,the target in the field of view in a complete driving track.In current research,the moving target detection and tracking algorithm based on binocular vision is studied by using UAV target tracking system as the research object.Firstly,according to the binocular vision system,the binocular camera is calibrated by selecting the calibration method with the accuracy and flexibility,and the calibration principle and the calculation process are analyzed.The internal and external parameters of the camera are obtained,and the output image is corrected.For the real-time tracking of UAV,the left and right images are matched by the BM stereo matching algorithm with fast speed,and the image with depth information is obtained.Secondly,based on the analysis of the background motion compensation algorithm,the SURF algorithm is used to detect the feature points of the background.Then,the background image is affine transformed.Finally,aiming at the interference of the light changes and the disturbance of the leaves during the target detection process,a three-frame difference algorithm combining the depth image and the gray image is adopted in the moving target detection process,which increases the accuracy of the moving target detection and provides accuracy target template for the subsequent tracking algorithms.Finally,the Camshift tracking algorithm is used to detect the moving target,and the Kalman filter algorithm is used to predict the target motion state.The problem of severe occlusion and large area background interference in the tracking process is analyzed.The Kalman filter and Camshift algorithm are improved to solve the above problems and enhance the robustness of the system.Through the construction of binocular vision platform,this paper presents a variety of algorithms using Matlab or C ++ language in the computer environment simulation and debugging completed,the use of video images for a large number of simulation experiments.