Moving Objects Tracking Algorithm Research Based On The Particle Filter

Moving Objects Tracking is of vital importance in both civil and military use. Demand of researching and developing tracking algorithms and systems with high reliability and high real time have been more and more in these year. Because vision procedure is actually Non-linear/non-Gaussian, using the corresponding techniques become a kind of important research tend in image fields. Particle filter is one of theories as description above, and has distinguishing features. Particle filter algorithm is main content of this paper, which can be implemented by a Bayesian recursion process though a Monte Carlo simulation method. The method is suitable for any non-linear system that could be represented with state model, It is more practical than conventional Kalman filter and its precision could approach optimal estimation. Particle filter is flexible and easy to be implemented.The core of this paper is to study the theory of particle filter, and based on particle filter the tracking of moving objects.Based on MS calculated the weight of the particle filter to track moving objects is one of the focuses of this paper, and it is also the fundamental work. The method is used to solve tracking problem and the tracking framework is formed accordingly. It will be an organic combination of a more rational way, MS calculated weight particle filter and through the distribution of particles of the weight, contribute to the back of a more accurate particle sampling and re-sampling, the method to further improve the particle filter to track the performance, achieve better moving target objects.Finally, the introduction of a new algorithm, and Kalman will be combined with particle filter for tracking moving objects. The Kalman filter calculate fast and properly, but it asks the object has linear-Gaussian moving characteristic; the particle filter can solve the problem in non-linear and non-Gaussian system, but the requirements of computational capability and storage capacity are higher than Kalman filter. When objects are non-occluded, moving objects'state recursion meet the linear-Gaussian condition, so we employ the Kalman filter. On the contrary, it is more accurate to model the moving objects' state recursion using particle filter. Thus, this paper combined above-mentioned two methods, Under the non-occluded situation, we adopt the connected-component tracker based on Kalman filter, while under the opposite situation, adopt the particle filterer based on MS calculated the weight.In experiments, we use the video screened by ourselves and PETS2006's testing video, the experimental result shows:As for the moving object tracking algorithm mentioned in the paper, based on MS calculated the weight of the particle filter to track moving objects can be achieved on the part of the occluded of tracking moving objects; in the situation of non-occluded and short-time occluded, it can implement accurate tracking. The algorithm is practical, real-time and robustness.