Shape Estimation Based Multiple Extended Target Tracking Algorithm

Due to the improvement of modern sensor resolution technology,each target occupies multiple resolution units of the sensor in the monitoring process,and each sampling unit produces multiple measurements,the target is called an extended target.At this time,the direct use of point target model to model the extended target,it is difficult to obtain the complete information of the target,and can give full play to the advantages of high-resolution sensors.Therefore,the idea of modeling the target as an extended target was born.However,the data association problem becomes more challenging due to the "one-to-many" correspondence between the extended target and the measurement.Then,the theory of Random Finite Set(RFS)was proposed to provide a new solution to the data association problem in extended targets,which has led to the continuous development of extended target tracking algorithms.In this paper,the RFS theory is used as the background for the Potential Balanced Multi-Extended Target Multi-Target Multi-Bernoulli(ETCBMeMBer)filter and the Poisson Multi-Bernoulli(PMB)filters in the form-state problem,and carry out research on ETCBMeMBer filtering based on spatio-temporal Gaussian processes and PMB filtering with derivatives,as follows:In this study,the RFS theory is used as the background for the potential balanced Multi-Extended Target Multi-Bernoulli(ET-CBMeMBer)filter and the Poisson Multi-Bernoulli(PMB)filters in the extended state problem and conduct research on ET-CBMeMBer filtering based on Spatio-Temporal Gaussian processes and PMB filtering with derivatives as follows:1.To address the tracking problem of irregular multi extended targets,this study proposes a potential multi extended target Balance Multi-Bernoulli(STGP-ETCBMeMBer)filtering algorithm based on the Spatio-Temporal Gaussian Process(STGP)model.First,the kinematic and shape states are expanded to the target states,which are modeled as a multi-Bernoulli RFS,and the STGP method is used to model the shape of the star-convex extended target to improve the accuracy of the algorithm for the shape estimation of the extended target.After that,the Gaussian Mixture(GM)implementation of the STGP-ETCBMeMBer filtering algorithm is derived by assuming that multiple likelihood functions corresponding to the same target measurement subset obey Gaussian distribution in the algorithm update stage.Finally,the performance of the STGP-ETCBMeMBer algorithm is verified through simulation comparison experiments,and the simulation results show that the algorithm has more accurate results in the shape estimation of extended targets.2.The standard Poisson multi-Bernoulli(PMB)filter for extended targets can hardly track spawning targets effectively.To resolve this problem,this study proposes an improved PMB tracking algorithm.The algorithm uses a multi-hypothesis model to predict the derived events in the filter prediction stage,and uses the random matrix method to model the original and spawning target shapes as ellipses with the state described as Gamma Gaussian inverse Wishart(GGIW)distribution.Finally,in the filter update stage,the prediction components are updated to obtain the motion state and expansion state estimation of the extended target.The simulation results show that the proposed algorithm effectively improves the tracking performance of the derived extended target compared with the standard PMB filtering algorithm.