Multi-Target Tracking Algorithm Research On Multi-Bernoulli Filtering

Multi-Target tracking technique is an important research in the field of information fusion, and has been widely used in military field. Random finite set(RFS) theory is introduced into multi-target tracking technology by Mahler, which can avoid the combinatorial explosion caused by data association algorithm, and open up new areas for multi-target tracking theory. In the RFS framework, multi-target multi-Bernoulli(Me MBer) filtering algorithm is proposed by Mahler. In order to solve the over-estimation of the targets’ number in Me MBer filtering algorithm, cardinality balanced Me MBer(CBMe MBer) filtering algorithm is proposed by Vo. With the increasing complexity tracking environment and rising target tracking precision, the original multi-target tracking theory also needs further study. This paper focuses on multi-target tracking problem under the relevant conditions based on CBMe MBer filtering algorithm.Firstly, Gaussian mixture CBMe MBer(GM-CBMe MBer) tracking algorithm of multiple maneuvering targets has been researched. Since the motion model of standard GM-CBMe MBer filter is not suitable for the motion state of maneuvering targets. In this paper a multiple maneuvering targets tracking algorithm is proposed by combining the GM-CBMe MBer filtering with the adaptive current statistical model. The new algorithm based on CBMe MBer improves the target motion state model in the prediction stage, and can adapt to changes in maneuvering target motion states. The simulation results show that the new algorithm can adapt to the changes of maneuvering target motion state, and make precise tracking.Secondly, this paper studies GM-CBMe MBer filtering algorithm under low detection rate. In this case, GM-CBMe MBer filtering algorithm has serious tracking loss because there is no enough measurement during the update process. To solve the problem, an improved GM-CBMe MBer filter is proposed in this paper. Gaussian terms with high weights are certainly processed in the original filtering processing, Simulation results show that the new algorithm can improve the tracking accuracy.