Research On The Multi-target Tracking Method Based On The Particle Bayesian Filter

The major function of multi-target tracking is not only to estimate the state of each target but also the taget number in the presence of clutter and noise.Nowadays,the multi-target tracking technique has been widely applied to military and civilian fields,therefore it is of important research significance.The marginal distribution Bayes(MDB)filter and sequential multi-target Bayes(SMB)filter are efficient for multi-target tracking.However,these two filters usually assume that the observation noise is a Gaussian noise,therefore they are inapplicable to the case of non-Gaussian noise.To extend the applicable scope of the MDB filter and SMB filter,we apply the particle filtering technique to the MDB filter and SMB filter,respectively;and propose the particle MDB filter and particle SMB filter in this thesis.The main content of the thesis is summarized as follows:1)We introduce the multi-target tracking Bayes filter based on the random finite set statistics,the probability hypothesis density(PHD)filter,MDB filter,SMB filter,and particle filtering technique.2)Applying the particle filtering method to the MDB filter,we propose a particle MDB filter.Based on the proposed particle MDB filter,we then develop two tracking algorithms for non-linear Gaussian system and non-linear non-Gaussian system,respectively.The simulation results show that the proposed particle MDB filter has a better multi-target tracking performance than the available filter in case of clutter interference,low detection rate,noise interference,and unknow target number.3)Applying the particle filtering method to the SMB filter,we also propose a particle SMB filter.Based on the proposed particle SMB filter,we then develop two implementations of this filter for non-linear Gaussian system and non-linear non-Gaussian system,respectively.The simulation results demonstrate that the proposed particle SMB filter is more efficient and accurate for target tracking than the available filter in the presence of clutter interference,noise interference,and unknow target number.