Research On Video Object Tracking Algorithm Based On Particle Filter

Video object tracking is an important research direction in the filed of computer vision. In the non-linear, non-Gaussian systems, so it can be applied to the field of video object tracking. Particle filter algorithm can still maintain good and stable performance characteristics. This paper focuses on particle filter. Target for video features, it can be divided into goals and gray color targets, the article focus on the study of these two different objectives tracking.In color targets, this paper targets the color histogram, which focuses on the target fast movement, part of the block, rotate, and zoom scale under the conditions of tracking problem. For the large amount of sample and low efficiency problem in particle filter algorithm, this paper proposed a Camshift optimized Particle filter Tracking Algorithm. The algorithm combines the particle filter algorithm and Camshift algorithm advantages. Camshift can make the particles aftered the transfer fastly move to the target area, so sample particles concentrated in the vicinity of the target to improve sampling efficiency and reduce the computational complexity.At the same time, For the similar color interference problems, the algorithm is to further processing, and Proposing separate treatment strategies. when the target similar to the color interference with surrounding objects is small, to take depending on the target percentage of the surrounding background of similar color Camshift optimization involved in adaptive.When more similar color disruptors, we will take a fusion target color and edge histogram method to calculate the particle weights. This method is only used in much similar color objects, so the algorithm can not only save time but also solve similar color interference, increasing the stability of the algorithm. The experimental results show that the algorithm has good real-time performance and robustness.For the gray video target tracking, this paper presents a new template match method in the framework of particle filter. This approach combines the correlation-matching method with particle filter. For the gray distribution features can only be adapted with the goal of small angle rotation and the small-scale changes in posture, active contour model proposed to update algorithm. During the template refreshing of the correlation algorithm, this paper applies the active model contours algorithm to get access to the object contours to reach the accurate object area so as to refresh the template exactly. Experimental results also demonstrate this method can track objects when object part-occlusion and various sharps scales.