A Study Of Multi Stealth Target Tracking Based On Random Finite Set Theory

Recently,approaches to mutli-target tracking based on random finite set(RFS)theory have been widely applied in military and civilian fields,and a variety of relevant technology has been developped due to the study of this problem.In modern fight scenario,stealth aircrafts with low observable,supersonic and maneuverable abilities are becoming main combat weapons among US-led western countries.But tracking performance of traditional filtering may decrease sharply when dealing with stealth targets.This thesis is aimed at extending the application of filtering according to the characteristics of stealth targets.In other words,we propose improved filter algorithms to increase their reliability and tracking accuracy.The major contributions are outlined as follows:Firstly,the theory of RFS is introduced and a description of some existing RFS based algorithms is made.Also,we give some comparison and analysis through several standard filters,and the results demonstrate the advantages of multi-target multi-Bernoulli(MeMBer)filter.Secondly,to tackle the stealth target tracking system with unknown detection parameters such as measurement variances and detection probability,we extend the traditional filter and propose a novel MeMBer filter algorithm based on variational Bayesian(VB)approximations.Also,a Beta Gaussian Inverse-Gamma mixture(BGIGM)implementation is made to estimate the Bernoulli parameter set and derive recursive equations of the filter.The results show that the proposed VB-MeMBer filter has more accurate cardinality estimation and smaller errors of optimal subpattern assignment(OSPA).Meanwhile,because the information of stealth targets would be intensively lost after threshold detection,we propose a MeMBer track before detect(TBD)scheme to deal with weak stealth targets.Also,we consider the complicated fluctuation features and incorporate the target fluctuation model with likelihood function.The results show that proposed algorithm is capable of estimating target kinematic states under low signal to noise ratio condition.And tracking performance has been improved due to the reconstruction of target fluctuation model.Finally,the multiple model approach is involved to tackle maneuvering targets.We augment the constant acceleration(CA)model into the model set to describe targets' acceleration and deceleration.The model set in this paper is able to simulate complex maneuvering situation and the results demonstrate that the proposed algorithm can deal with multiple maneuvering target tracking.