Moving Objects Tracking Based-on Particle Filter Algorithm

With the rapid development of computer technology, the arrival of the era of intelligent information, needs for intelligent video surveillance system is growing in the field of security, transportation, military and other fields. Video target tracking is one of the core content of the intelligent video surveillance, which combines the advanced technology in many areas such as image processing, pattern recognition, artificial intelligence and computer, has been widely applied in many aspects of military visual guidance, safety monitoring, traffic management, medical diagnostics and meteorological analysis. In recent years, a large number of domestic and foreign scholars have important study on the video object tracking, and made certain achievements in scientific research. As a filtering algorithm, particle filter because of the nonlinear and non-Gaussian system adaptability, such as target tracking, status monitoring, fault detection, and other areas of computer vision has its unique advantages, and has been widely studied. However, many problems which are large amount of calculation and real-time differential still persist in particle filter algorithm. To solve the above problems, based on particle filter algorithm for video object tracking is studied in this thesis, to improve the robustness and real-time of the target tracking. In this thesis, the work is done as follows:(1)With the instability problems of target model in particle filter, a target model with combination of color statistical characteristics, target characteristics and color gradient projection feature is proposed, and applied it to target tracking to improve the robustness of tracking.(2)In order to improve the accuracy of particle filter tracking, the modeling of the state transition equation is improved for Elementary particle filter algorithm. Then for the disadvantages of high complexity in particle filter algorithm, the improved particle filter algorithm is optimized relevantly. So it improves the efficiency of operation of the algorithm.(3)Based on the particle filter algorithm, the above proposed algorithm and the improved algorithm, the video sequence in the video file has been experimented by target tracking simulation. The experimental results show that the algorithm has better robustness and reliability, to verify the effectiveness of the proposed algorithm.