The Research On Multi-extended Target Tracking Algorithm Based On Multi-bernoulli Filter

With the rapid development of sensor technology and the increasing complexity of the target tracking task,it is increasingly popular that one target can occupy multiple sensor resolution cells.The extended target tracking has attracted more and more attention from domestic and foreign scholars.In many practical applications,there is an urgent need to track multiple extended targets.Due to an extended may correspond to multiple measurements,the association between target and measurement is more ambiguous,and the multi-target tracking becomes a more difficult problem to solve.In recent years,multi-target tracking based on random finite set(RFS)has been widely concerned because it is unnecessary to make a complex data association.Based on Cardinality balanced multi-target multi-Bernoulli filter(CBMeMBerF),this thesis studies the tracking of multiple extended target with irregular shape in the theoretical framework of RFS,more detailed works are as follows:(1)Considering the tracking of multi-extended target with irregular shape in complicated and uncertain environment,a multi-extended target multi-Bernoulli algorithm based on random hypersurface model(RHM)is proposed.First,in the framework of finite set statistics(FISST),the multi-Bernoulli RFS and Poisson RFS are used to model multi-extended target state and measurement respectively.Subsequently,in order to describe the relationship between extended target state(including the kinematic parameters and the irregular shape parameters)and measurement,the pseudo measurement equation based on RHM is established.And then,the detailed Gaussian mixture implementation of multi-extended target multi-Bernoulli filter based on RHM is derived.Besides,this paper establishes a metric to evaluate the irregular shape estimation performance of multi-extended target tracking.Finally,the simulation results show the effectiveness of the proposed approach applied in the multi-extended target and multi-group target tracking.(2)Based on Gaussian process regression(GPR)method,the star-convex extended target measurement model is established,and the multi-extended target tracking algorithm based on GPR is proposed.Unlike the star-convex RHM,which uses the Fourier series description of radial function,the GPR models the star-convex object through the mean function and covariance function of radial function.First,this chapter analyzes the theory of Gaussian process(GP)and GPR in detail,and a nonlinear measurement model based on GPR is established.And then,using the extended Kalman filtering to calculate the likelihood function,a CBMeMBer filtering algorithm for multi-extended target tracking based on GP is proposed,and its' Gaussian mixture implementation process is derived.Finally,the simulation results show that the proposed algorithm can accurately estimate the shape,kinematic state and target number of multiple extended targets.Furthermore,the effectiveness of the algorithm is verified in the tracking of multiple star-convex extended targets.