Multiple target tracking using the extended Kalman particle probability hypothesis density filter

The Particle Probability Hypothesis Density Filter (PFPHD) provides a numeric solution for the probability hypothesis density (PHD) filter, which propagates the first-order moment of the multi-target posterior instead of the posterior distribution itself because evaluating the multiple-target posterior distribution is currently computationally intractable. The PFPHD considers the target states as a single global target state and then avoids data association steps. Various implementations using particle filter had shown the efficiency of this method in real time applications. However, most of them use the state transition prior as the proposal distribution to draw particles from. Because the state transition does not take into account the most recent observation, we present, in this paper, a new approach that mixes the PFPHD filter with the Extended Kalman filter (EKF) named EK-PFPHD filter. The first part provides the general probabilistic framework to handle non linear non gaussian systems when the second part generates better proposal distributions by considering the updated observation. Simulation shows that the proposed filter outperforms the PFPHD filter.