Research On Target Tracking Based On Binocular Vision

With the development of science and technology,the application of computer vision is more and more extensive in our lives.Object tracking is a very important part in computer vision technology.The performance of the existing target tracking algorithm in complex environment is not ideal.Therefore,this paper proposes a tracking algorithm based on binocular stereo vision.The main work of this paper is as follows:Firstly,a stereo image pair is obtained by using a binocular stereo camera.Then the disparity map is calculated according to the stereo image.Then the 3D depth map is obtained based on binocular vision system parameters..When the camera angle of binocular stereo vision system reaches a certain value,there is a big error in the disparity map.This is because there are some errors in the calibration of the system parameters.In this paper,a method is proposed to modify the rotation parameters by decomposing the rotation parameters between the left and right cameras.In order to obtain the stability of the tracking and the simplicity of the algorithm,this paper selects the HOG edge feature and the Lab color histogram of the target to record the edge feature and color feature of the target.Setting the original target and the target variable of the previous frame.The Kalman filter is used to help the finding of the target's position.Finally,in order to ensure the stability of the target tracking in the case of occlusion.In this paper,a mapping image from target's candidates is produce,and then the occlusion of the target is determined to avoid the error tracking.If the target is detected,the Lab color feature of the original target is updated according to the similarity.