A Robust Unscented Kalman Filter for Tracking Star-Convex Extended Target with the Non-uniform Scattering Sources

Conventional extended target tracking algorithms rely upon the hypothesis of the uniformly distributed scattering sources. However, the change of radar azimuth makes the statistics characteristic of the scattering source changed. In this paper, a robust star-convex extended target tracking algorithm is proposed to deal with this issue. Uniform distributed pseudo-measurement is generated by using the Centroid Contour Distance method (CCD) and Acceptance-Rejection sampling method, and then Double Layer Unscented Kalman filter (UKF) is derived to calculate the posterior intensity of kinematic state and shape parameters. Simulation results confirm the superiority of the proposed methods in comparison with classic UKF.