An improvement on GM-PHD filter for occluded target tracking

The Probability Hypothesis Density (PHD) filter is the first-order momentum of Bayesian multi-target filter. The Gaussian Mixture PHD (GM-PHD) implementation is a closed form solution for the PHD filter. When targets are too close to each other, such as occlusion condition, the performance of the GM-PHD filter degrades significantly. In this paper a novel algorithm is proposed to improve this drawback. Our method employs a renormalization scheme to re-manage the weights assigned to each target in the GM-PHD recursion. Simulation results show that our proposed approach significantly improves the overall estimation performance of GM-PHD filter.