Research And Implementation Of Video Sequence Target Tracking Algorithm Based On Particle Filter

Moving object tracking technology in video sequences is one of the most active subjects in the field of computer vision technology. It relates to image processing, pattern recognition, automatic control and many other fields with strong research value. Based on meanshift theory, Camshift is a good tracking algorithm. As a simple algorithm with strong real-time performance and robustness, and it can basically track the moving objects accurately in video sequences under the circumstance of the target is obscured. Therefore it has been widespread concerned in the field of target tracking.The basic mean shift algorithm is able to handle simple target tracking problem, but there exists some problems of high computational complexity, long processing time and easily missing target. For target tracking in video sequences, this paper introduces the theory of particle filtering. First of all, according to the former image to manually determine tarcking target, and also according to the movement state of former several frame to get the system motion equation and system observation equation, and let the fast tracking of moving target can be discussed within the constraints of inside the video sequence, Then based on the related theory and definition of particle filter, the video sequence is sampled ty the mean and variance of normal distribution to calculate the weight value of particle. Then, this paper is committed to improve particle filtering algorithm and propose a new resampling algorithm to predict centroid position of target. By theoretical analysis and experimental verification, it can be known that the particle filter algorithm has good adaptability in the field of video sequence’s tracking target. In the condition of target is obscured, combing the mean shift offset and particle filter can correctly track the target’s physical location. Next, this paper finishes the three-dimension reconstruction of target tracking model, and analyze the binocular stereo vision model in the context of camera calibration in order to achieve the purpose of time synchronization by adjusting video. Finally, the target tracking system of binocular stereo vision is designed and implemented, and the regular motion target, target is obscured, not regular motion target are also analyzed and also experiments were carried out.The experimental results shows that the improved target tracking algorithm based on particle filter can well solve the problem of tracking video sequence’s moving target, and reduce the complexity of the algorithm, improve the efficiency of computation. It can successfully track moving target and handle the condition of target is obscured at the expense of less processing time. The experiment results also prove that the particle filter is great effective for the improvement of video sequence tracking algorithm.