Research On Multi-extended Targets Tracking Algorithm Based On Random Finite Sets And Labeled Particles

With the development of radar and computer sicience,target tracking technology has developed rapidly,and it is widely used in military and civilian fields.Extended targets tracking is very important.The target tracking methods based on random finite sets(RFS)do not require data association between target states and measurements,avoiding combinatorial explosion.So it is widely used in extended targets tracking.Therefore,extended targets tracking algorithms based on RFS are discussed in this paper.The main work are as follows:Firstly,current multi-extended targets tracking methods based on elliptical random hypersurface(RHM)can't distinguish different targets.So ET-LP-PHD filtering algorithm based on elliptical RHM and labeled particles is proposed in this paper.A new labeled particles supplement method is used in the prediction step of the extended target tracking.And label update processing is added after the measurement update step.So the proposed algorithm can distinguish different targets.The experiment is conducted by simulating a twodimensional scene,and the proposed algorithm is compared with the ET-SMC-PHD filtering algorithm.The experiment shows that the ET-LP-PHD filtering algorithm can not only effectively track the centroid and extended shape of multiple extended targets,but also can obtain the trajectories of multiple extended targets.Subsequently,considering the elliptical maneuvering extended targets tracking,the ETIMM-LP-PHD filtering algorithm based on elliptical RHM is proposed.Several maneuvering target tracking models,intersection over union(IOU)and its optimization are introduced together with the ET-IMM-SMC-PHD filtering.The ET-IMM-LP-PHD filtering algorithm for multi-maneuvering extended targets tracking is proposed on the basis of the ET-LP-PHD algorithm and interacting multiple model.A simultion is carried out to campare the accuracy of the ET-IMM-SMC-PHD algorithm and the ET-IMM-LP-PHD algorithm.The experiment shows that the proposed algorithm can obtain the extended target trajectories to distinguish different targets while estimating target shape and centroid accurately.Finally,the ET-IMM-LBP-PHD algorithm is proposed on the basis of BP-PHD filtering and interactive multiple models.Simulation experiment shows that the algorithm can not only estimate the centroid states of multi-extended targets accurately in a high-noise environment,but also obtain multi-target trajectories.So it can distinguish different targets.