Realtime image sequence interpretation for video-surveillance applications

This paper describes a generic framework for the realtime interpretation of video-surveillance scenes. The tool we have built according to this framework performs the understanding process up to the online recognition of ongoing scenario with respect to a library of scenario models dedicated to a particular application. This framework is currently applied for the European Esprit project PASSWORDS, the aim of which is to build a complete system for the automatic analysis and interpretation of video-surveillance image sequences.