Research On Video Target Tracking Technology Based On Mean Shift

Video object tracking(VOT)is one of the important research contents in the field of computer vision.With the rapid development of computer technology,image processing technology and artificial intelligence technology,it has been widely used in many fields,such as intelligent monitoring,intelligent transportation and medical diagnosis and treatment.But the actual tracking environment is complex,such as the change of illumination intensity,the change of the shape and size of the object itself,and the in-plane and out-of-plane rotation will lead to the performance of the object tracking technology algorithm is reduced.Therefore,improving the robustness,accuracy and real-time of video object tracking has become an important research content of its key technology.Video object tracking technology can be divided into object detection and object tracking.Firstly,based on the accuracy of detection,this paper studies the traditional video object detection algorithm,analyzes the advantages and disadvantages of inter-frame difference method,background subtraction method and optical flow method,and deeply analyzes several classical feature point matching methods,so as to combine the Oriented Fast and Rotated BRIEF(ORB)feature point matching method with the inter-frame difference method,and proposes an improved inter-frame difference method based on feature point matching to provide accurate initial values for subsequent object tracking.Secondly,aiming at the vulnerability of the traditional Mean Shift algorithm to complex background,this paper proposes a background weighted multi-feature fusion Mean Shift algorithm(BWMFMS).In the traditional Mean Shift algorithm,the background weighting coefficient is introduced,and the texture feature and edge feature are integrated to establish a precise object model.The object model is scaled to solve the problem of "small scale loitering" and "scale infinite growth" in object tracking.The proposed algorithm improves the robustness and accuracy of the tracking algorithm.Finally,aiming at the traditional Mean Shift algorithm object occlusion drift problem and the problem of unable to track fast moving objects,and the Cubature Kalman Filter(CKF)is fused with the BWMFMS algorithm.In this paper,the CKF-BWMFMS algorithm is proposed.And the predicted value of Cubature Kalman Filter,as the initial value of the BWMFMS algorithm.BWMFMS algorithm results are taken as the measured value of the Cubature Kalman Filter,and finally the position of the object is obtained.The algorithm can guarantee the accuracy and real-time performance,improve the robustness,and solve the problem of occlusion drift and the inability to track fast moving objects.