Feature aided Gaussian mixture probability hypothesis density filter with modified 2D assignment

In order to track multiple targets with time-varying number of targets, the paper proposed a new feature aided Gaussian mixture probability hypothesis density (FA-GM-PHD) filter, and adopts a modified 2-D assignment algorithm to carry out the data association and manage the tracks in the FA-GM-PHD filter. The target feature information incorporated into the FA-GM-PHD filter is target Doppler and target down-range extent. With two typical multi-target tracking scenarios, the simulation results in the paper have verified that the FA-GM-PHD filter has much higher correct data association probability and filtering precision of target states than GM-PHD, and it can estimate the number of targets more stably and precisely than GM-PHD. The main shortcoming of FA-GM-PHD is that it has delayed estimate of target´s number at the spawning time than GM-PHD, which will be studied in the future work.