Research On Multisensor Multitarget Tracking Algorithm Based On Belief Propagation

The belief propagation algorithm is for solving Markov random field.This paper will establish the Markov random field model for multi-sensor and multi target tracking,and the belief propagation algorithm on this model will be provided.Based upon the facts provided before,this paper will present both belief propagation algorithm based on particle and belief propagation algorithm based on Gaussian mixture,which are both numerical methods.In order to break the restriction of belief propagation on limiting number for target tracking,this paper not only presents variable structure algorithm,but also gives three kinds of variable structure specific strategies,namely,variable structure strategy based on measurement,variable structure strategy based on joint initiation and strategy based on heuristic.Finally,this paper will present an algorithm based on both random finite set statistical theory and belief propagation algorithm,which is also called belief propagation probability hypothesis density filter.The simulation will show that the multi target tracking algorithm based on belief propagation algorithm can accomplish the tracking task when located in the noisy environment,and with high accuracy.The multi target tracking algorithm based on variable structure belief propagation can adaptively adjust the set of potential targets according to the number of actual targets.Moreover,the variable structure strategy based on heuristic logic is slightly better than the strategy based on joint initiation,and the performance of them is much better than that based on measurement.The belief propagation probability hypothesis density method is more efficient than the belief propagation algorithm.Comparing to probability hypothesis density filter,it is much better on tracking accuracy.