Multi-Target Tracking and Behavior Analysis Method for Video Surveillance Applications

Abstract

In order to obtain a satisfactory performance of visual tracking and video surveillance in complex dynamic scenes without the supervision of qualified workers, an efficient visual detection and tracking method is proposed, which can realize target counting and surveillance, behavior analysis and abnormal detection. Multitargets tracking method based on novel Bayesian tracking model can manage multimodal distributions without explicitly computing the association between tracked targets and detections. The proposed algorithm is compared with recent works, which shows that it is robust to erroneous, distorted and missing detections and it can be applied in security and management of access points.