Automatic detection of abnormal human behavior using video processing

S.A. Velastin

Kingston University, Kingston upon Thames, UK

Closed-circuit television (CCTV) systems are regarded as a major tool for surveillance within urban environments, capable of monitoring wide areas to identify and track individuals, detect potentially criminal or terrorist behavior, locate intruders in secure areas, and provide various kinds of situation awareness. The socio-economic context is that in a country like the United Kingdom, for example, the annual cost associated with anti-social behavior alone has been estimated at around € 5000 million.

The level of deployment of CCTV in urban environments continues to grow, and there are indicators that they can be perceived by the public at large as having a positive benefit. Such con.dence in their operation can only be maintained if the vast amounts of data generated by large numbers of cameras can be effectively and efficiently analyzed. Manual observation of many cameras is highly labor-intensive, and imposes unrealistic demands on the attention capacity of operators and as a consequence the majority of live images are un-observed, functioning mainly to support post-event inspection of recorded video (another labor-intensive task).

What is therefore desirable to support efforts to maintain an acceptable balance between the rights to personal security and privacy, is the availability of computer-based systems that can be capable of identifying potentially dangerous or suspicious behavior so as to timely inform those who might be in a position to prevent or control such situations. A particular challenge is to demonstrate robust systems that can operate continuously with little human intervention in fairly complex scenarios (e.g. public transport, retailing facilities, streets, etc.) and that also scale up so as to have the large geographical coverage required in practical applications.

This symposium will give attendees the opportunity to understand what is currently possible with computer vision technologies, what the main challenges are and discuss both the technical and social implications of the deployment of these systems.

Paper presented at Measuring Behavior 2005 , 5^th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.

© 2005 Noldus Information Technology bv