Occlusion Handling for Pedestrian Tracking Using Partial Object Template-based Component Particle Filter

Pedestrian tracking plays a crucial role in security and intelligent video surveillance. Occlusion handling is a challenging concern in tracking multiple people. Adaptive, advanced solutions are required to accurately track pedestrians for video surveillance purposes. This paper presents a novel method of tracking multiple people in occlusion conditions. In this study, we developed a combined componentbased human-shaped template and a particle filter, resulting in improved object occlusion handling. The proposed system is capable of tracking specific people in real-time, as well as handling the occlusion of multiple objects. We predicted occlusions using the Kalman filter, and tracked objects continuously using our component-shape-template particle filter. The experimental results show that our algorithm is feasible and stable. In tests, the proposed tracking algorithm achieved an accuracy of up to 99.7%. The proposed approach outperformed that of comparable methods in analyzing all test video datasets in this study. Our low false-negative rate demonstrates that the proposed tracking method is robust and offers superior occlusion handling.