An improved PHD filter based on variational Bayesian method for multi-target tracking

This paper presents an improved probability hypothesis density (PHD) filter for the multi-target tracking scenarios with unknown measurement noise variances. By introducing the variational Bayesian (VB) method into the PHD recursion, not only the states and number of targets, but also the measurement noise variances can be jointly estimated. Moreover, a closed-form solution to the improved PHD filter for linear Gaussian multi-target model is derived using inverse Gamma and Gaussian mixtures. Simulation results demonstrate the effectiveness of the proposed algorithm for the multi-target tracking scenarios with unknown measurement noise variances.