Multi-Features Particle PHD Filtering for Multiple Humans Tracking

This paper proposes multi-features visual tracking algorithm based on the particle Probability Hypothesis Density filter, which allows accurate and robust tracking under the circumstance of visual tracking. We apply a particle PHD filter implementation to the multiple humans tracking using multi-features observation that exploits skin and head-and-shoulder boundary as its prior density. The relevance of our approach to the problem of multiple humans tracking is then investigated using a tracker which is able to follow the state according to the humans' motion. The accuracy and robustness are evaluated and compared using real visual tracking experiments.