Competitive Gaussian mixture probability hypothesis density filter for multiple target tracking in the presence of ambiguity and occlusion

Abstract

The Gaussian mixture probability hypothesis density (GM-PHD) filter has recently been devised as a closed-form recursion for PHD filter for multiple target tracking. The GM-PHD filter works successfully when targets do not move near each other. However, the estimation performance of the GM-PHD filter degrades when targets are in close proximity, such as occlusion condition. In this study, the authors propose a novel approach to improve this drawback. The proposed method employs a renormalisation scheme to rearrange the weights assigned to each target in GM-PHD recursion. Simulation results achieved for different clutter rates and different probabilities of detection show that the proposed approach significantly improves the overall estimation performance compared with the original GM-PHD filter.