Tracking Of Moving Object Based On Particle Swarm Optimization And Kalman Filter

In the field of computer vision, tracking of moving object is one of the key issues, and its application fields include video surveillance, military, video encoding, transport and so on. The algorithm of moving target detection and tracking is one of the core technologies in many areas such as intelligent video tracking systems, the field of computer vision and so on. The researches of the algorithm have already made major progress. A lot of researchers have made a series of creative algorithms, but many difficult issues need to be analysed, researched and discussed. Aiming at the tracking inaccuracy, low real-time and poor treatment on the occlusion issue in common tracking algorithms of moving object, this paper presents a new tracking algorithm of moving object based on Particle Swarm Optimization (PSO) and Kalman filter.Firstly, this paper made a comprehensive overview of moving target detection and tracking. The basic theoretical knowledge and the specific methods used in this paper were analysed and described in detail. Grayscaling color image, denoise, histogram equalization, morphology and other aspects of related content were included in this part. Secondly, the paper briefly introduced and discussed several classical algorithms of moving target detection. This article also specifically addressed the background subtraction which was employed in this paper.Then the paper elaborated on the core of the paper. To begin with, the paper introduced the basic theoretical knowledge of Particle Swarm Optimization algorithm and Kalman filter. In addition, in order to solve the tracking problem better, we could combine Particle Swarm Optimization with the Kalman filter. The specific algorithm was described as follows:the possible position of moving target center in the next frame image was predicted by Kalman filter, which reduced the search scope greatly and set search region of target which was generated around the center position. Then with the gray statistical characteristics of the target template and the candidate regions matching, this method was used to ensure tracking accuracy. In order to improve accuracy and real-time, PSO was utilized to search the best area which was the most similar to the target template in the search region, as a result, the optimal center was found and the best position was used as an observed value of Kalman filter for next prediction. This is the specific method of Particle Swarm Optimization algorithm combined with Kalman filter.Finally, this paper carried on the experiment simulation by Matlab7.0.1. Three sections of the video test sequence, named WalkByShop1 front, OneLeaveShopReenter2cor and OneShopOneWait1 front, were from the CAVIAR data set. The experimental results show that the new method is effective and robust. It can handle partial occlusion and the interference of background better.