Research On Algorithms Of Multiple Extended Targets Tracking Based On Random Finite Set

With the development of sensor resolution, the assumption that a target is viewed as a point no longer holds. At this point, we treated this target as an extended target. Since each extended target will produce more than one measurement at each sampling period, if we associate measurement with the relevant target, there will be great difficulties. Therefore, the study of a more effective and real-time method is of great practical significance and value. In recent years, methods of extended target tracking based on random finite set have been widely recognized. Because this theory can handle some of the problems while traditional tracking methods can’t, and the computational complexity of which is much smaller than that of traditional algorithm.This paper focuses on the research on tracking algorithms based on random finite set, two aspects are included: 1)a Multiple extended target tracking method based on random matrix model. Including two representative algorithms: GIW-PHD and GGIW-CPHD filter.(2)a Multiple extended target tracking method based on random hypersurface model. The following algorithm is proposed: RHM-GGM-CPHD filter. The main contents are as follows:1. A gaussian inverse Wishart PHD filter for extended target tracking algorithm based on random matrix. The algorithm not only takes the target state but also the target extension into account.It models the target state as a Gaussian distribution while the target extension as an inverse Wishart distribution, so that the parameters of the Gaussian distribution and inverse Wishart distribution can be updated by measurements, in this way, we can track the position, size, direction information of a target. Based on the above statement, GGIW-CPHD algorithm can not only improve the performance of the PHD filter in target number estimation, but also consideres the measurement rate as a gamma distribution, by adding this parameter to the target state, the tracking performance will be improved.2.In view of the difficulty of estimating the shape of extended targets and the low accuracy in multiple extended target tracking in the presence of clutters, a Gamma Gaussian-Mixture cardinalized probability hypothesis density filter(RHM-GGM-CPHD)which can adaptive estimate the shape of extended targets is proposed. This algorithm is the same as GIW-PHD and GGIW-CPHD, considering the extension of target. Firstly, the extension of targets are modeled as ellipse random hypersurface model(RHM), and then it is embedded into CPHD filter. The extended targets are tracked by its centroid states, the shape and orentiation of reletated ellipse. Simulation for tracking of an unknown number of targets in the presence of clutter is made, which shows the proposed algorithm outperforms the Gamma Gaussian Inverse Wishart CPHD filter based on the random matrix in estimation of extension’s major and minor axis, as well as centroid states.