Semantic interpretation of object activities in a surveillance system

Activity analysis and semantic interpretation of tracked targets in a dynamic image sequence has recently attracted more attentions in computer vision. In this paper, a framework for semantic interpretation of vehicle and pedestrian behaviors is proposed for practical applications in visual traffic surveillance. The trajectories recorded in the visual tracking process are analyzed using dynamic clustering and classification on which high level semantic interpretation is based Experimental results are presented to illustrate the performance of the proposed algorithm.