A kernel particle filter algorithm for joint tracking and classification

For radar surveillance system, target tracking and classification are two major functions. A kernel particle filter approach with improved information mutual feedback is presented for joint tracking and classification. Delay, Doppler and Radar cross section measurements are used to estimate target state and class respectively. It invokes the kernel particle filter and point model for nonlinear estimation with less amount of calculation. Mutual feedback structure is used to improve the classification probability and estimation accuracy. Simulation results show the efficiency of the proposed method.