Research Of Moving Object Tracking Based On Video Images Sequence

Moving object tracking in image sequence is a hot research focus in the field of computer vision and image processing. It is widely used in military guidance, video surveillance, intelligent transportation, moving image compression coding and other fields. Therefore, the study of moving object tracking algorithm has a very important practical meaning.Moving object tracking process consists of two parts: first, the detection and extraction of the target to determine the target position in the image sequence; then the association with the target position in successive image frames, which is used to complete the target location thus find the target trajectory.The main content of this paper is analysis and implementation to the moving object tracking method, and the research is mainly focused on the two different kinds of object tracking method, and solving the different problems appeared in the process of tracking effectively. The main contributions of this paper are summarized as follows:1. Moving object detection and extraction, which is the premise and foundation of object tracking tasks, is directly related to the accuracy of object tracking. In this paper, we conducted experimental analysis of various detection algorithms including background difference method, the interframe difference method and optical flow method, and also pointed out these algorithms’ applicable scope and advantages and disadvantages.2. In the study of object tracking methods, after describing of the object tracking process and its classification, we analyzed and completed several common object tracking methods, such as Mean Shift algorithm, Kalman Filter algorithm and Particle Filter algorithm.3. In the case of mean shift algorithm, object is easily lost when the object is occluded or its scene is complex, because we can’t effectively predict the moving object state in these cases. In order to overcome this deficiency, we introduce object occlusion judgment and object model update mechanism after combining Mean Shift algorithm and Kalman Filter algorithm. The experimental results show that after improving the algorithm’s effectiveness and timeliness have been well improved.4. Finally, to solve the problem of particle degradation in particle filtering algorithm, we proposed a new method. We combine KLD-Sampling algorithm and Gaussian particle filter algorithm in the improved method, implementing the adaptability of particle population, thus avoiding the phenomenon of particle degradation and improving the Robustness and stability of the algorithm. Finally, we made comparison experiments between original algorithm and the improved algorithm, the result showing that after improving algorithm’s effectiveness have been well improved.