The Research Of Video Target Tracking Based On MeanShift Algorithm

With the development of science and technology, especially the rapid development of computer technology, the computer vision has become possible. Target tracking is a hot issue in the field of computer vision, widely used in video surveillance, intelligent traffic, robot vision navigation and so on. The target tracking system demands highly for the robustness and real-time performance, around which the mainstream Mean Shift algorithm for target tracking system is researched in this paper. The main contents are as follows:First of all, the background and current development of target tracking are summarized and the theoretical basis of Mean Shift algorithm in target tracking is analyzed to determine its merits and faults.Secondly, as the traditional Mean Shift tracking algorithm is based on the target color feature model, of which the color feature is global information, sometimes probably identical in accordance with different targets, the Mean Shift tracking algorithm based on the block color histogram is proposed. The algorithm uses the block method to fit the spatial position information into the target model, and the validity is verified through simulation experiments.Thirdly, with Mean Shift target tracking algorithm determining target position through comparing the similarity of the original target and the candidate model, it is not sufficient to describe the target according to the characteristics of a single color model and the background easily intervenes. Aimed at this problem, combined with color, edge and LTP texture features, computing the characteristic for the discrimination of target and background, a multiple feature fusion strategy is adopted in the target description, based on which a weight fusion is carried out subsequently. Through simulation and comparison between the two kinds of algorithms, the anti-interference ability of the improved algorithm is verified.Finally, for the problem of occlusion which often happens and is difficult to solve in the target tracking, the combination of the extended Kalman filter and the Mean Shift algorithm is proposed. Possible location of the target is predicted depending on the estimation ability of extended Kalman filter for the motion system, with that, the Mean Shift algorithm is used to improve the anti-occlusion ability of the algorithm. In order to verify the validity of the algorithm, the two algorithms are tested for occlusion by simulation experiment.