Target Tracking Based On Mean-shift Algorithm

As one of the key disciplines of intelligent video surveillance, video target tracking has a very broad application prospects, such as human-computer interaction, medical diagnostics, traffic control, Video surveillance and other aspects, which is involved in an important discipline of computer technology, signal processing, pattern recognition and computer vision, etc. In recent years, a large number of researchers studied for the subject in-depth and has made many achievements. Wherein, the mean-shift tracking algorithm has been widely used with high efficiency, simplicity and other advantages. Because of the target template is not robust enough, the tracking performance is poor when in the case of complex background and illumination changes. In order to improve the above problems, the focus of this thesis is to study the robust modeling of the target in mean shift algorithm, thus it is increasing the timeliness and accuracy of tracking. The work done in this thesis is as follows:1. For the process of tracking in the mean shift algorithm, since the target tracking performance is not ideal, even the problem of tracking failure, which is resulted as the target model is not stable enough, a new target modeling is proposed. Target model established by this method is more stable for the mean shift tracking algorithm can improve the accuracy and robustness of target tracking.2. In order to verify the performance of the proposed method established the target model, the target model is used for the mean shift tracking algorithm to be compared with experiments under different experimental conditions, which is established by the traditional gray methods, the Classic RGB method and the proposed method.3. When target color is rich, the modeling method in this thesis has good tracking result. Therefore, we put forward a method to determine the target color, and the feasibility of this method is verified through experimental example.