Towards Suspicious Behavior Discovery in Video Surveillance System

Video surveillance systems are becoming common in commercial, industrial, and residential environments. The systems in used are constructed mainly by hard devices with no or very few soft intelligence. It is difficult for human to recognize important events as they happening and to control over unwilling situations by staring at the screens all the time. Soft intelligence to identify human behaviors in the surveillance systems is expected. A systempsilas architecture for this goal is presented in this paper. Bottom-up processing methods and top-down design schemes are integrated in the architecture. The integration may increase the accuracy of relevance algorithms and reduce the computing cost. The feasibility of the system is assured.