Particle implementation of the multi-group multi-target probability hypothesis density filter for multi-group target tracking

We propose a particle implementation for the multi-group multi-target probability hypothesis density (MGMT-PHD) filter in this paper. It provides estimates of motion state of multi-group target centers as well as its components. The algorithm models multi-group centers as parent process, components as daughter processes related to centers. With separation of the two interacting point processes, the huge computational complexity arising from high-dimensional joint estimation is decreased. In the simulation scenario, we set a typical complicated multi-group target scene with target appearance and disappearance and tracks crossing to test the performance of the proposed algorithm.