Visual target tracking in occlusion condition: A GM-PHD-based approach

The Gaussian mixture probability hypothesis density (GM-PHD) filter has recently been proposed for multiple target tracking in the presence of some uncertainties including miss detection. However, the performance of this filter degrades in occlusion where miss detection occurs for a several consecutive frames. In this paper, we propose a novel approach to address this issue of GM-PHD filter. The proposed method estimates the probability of detecting of each target during tracking dynamically, and incorporates this information to cope with occlusion. The experimental results provided for real dataset as well as simulated dataset show that the suggested method improves the performance of GM-PHD for tracking video targets in occlusion.