Bearings-only tracking using probability hypothesis density filter in modified polar coordinates

Bearings-only tracking in the Cartesian coordinates (CC) presents the poor convergence and erratic behavior due to the variable observability. If considering the multi-target tracking problem and the influence of measurement origin uncertainty, this problem will be more challenging. This paper discusses a modified polar coordinates (MPC) based Gaussian mixture probability hypothesis density (PHD) filter. This approach can decouple the observable and unobservable components of the target state by using MPC and make the best of PHD to accomplish the complicated association problem. Moreover, we use the unscented Kalman extension of the GM-PHD filter to resolve the high nonlinear of the state equation in the MPC. The application of the proposed approach in simulation proves its effectiveness and practicability.