A CPHD Filter Based on EM Star-Convex Random Hypersurface Model for Multiple Extended Targets

Aiming at the problem of tracking multiple extended targets with irregular shapes in complex scenes, this paper presents a Cardinalized Probability Hypothesis Density (CPHD) filter built on EM iteration and Star-Convex Random Hypersurface Model (SRHM). Firstly, based on the theory of Finite Set Statistics (FISST), the Bayesian filtering framework for multiple extended targets is established by using the CPHD filter. Then, SRHM is adopted to describe the measurement source distribution of the star-convex extended target, and Unscented Transform is used to embed the CPHD filtering process. In addition, the incorporation of EM iteration significantly alleviates the dependence of the SRHM-CPHD filter on the RHM prior. The simulation results show that the tracking performance of the proposed EMSRHM-CPHD filter is better than that of the SRHM-CPHD filter, and the parameter estimation of the extended targets is more accurate.