The paper addresses the multi-target tracking problem in the framework of asynchronous radars system. Based on probability hypothesis density (PHD) filter, we propose two fusion methods. The first one transfers information between different radar nodes, named as Sequence Process PHD (SP-PHD) fusion, it combines the prediction and update steps of a PHD filter, time alignment and recursive state estimation based on sequential process fusion, and achieves overall good performance; the second one, namely Fixed-nodes PHD (FN-PHD) fusion, can achieve similar performance but with much lower communication cost and computational cost. Firstly, radars track targets by PHD filter, respectively. Then, choose the targets state closest to fusion node, predict and fuse all states together among radars based on generalized Covariance Intersection (GCI). We implement the two proposed ways using the Gaussian Mixture (GM) approximations and demonstrate their performance in simulation.
展开▼
