The general problem addressed in this paper is how to provide intelligence surveillance in an environment that consists of complex scenes. In an intelligent surveillance system, events of interest must be recognized and complex environments must be understood. Whereas much of the previous work in this area has been limited to a particular problem or a specific environment, the authors’ approach is intended to be adaptable to new events, different environments, and the addition of new monitors to the surveillance system.
In this approach, “intelligent agents cooperate to monitor the whole scenario so that each agent can focus on a particular event of interest (such as trajectories of moving objects, pedestrian crossings, and crowds).” They incorporate fuzzy logic into an approach called normality analysis of events, which means that their approach is intended to “estimate the normality degree of recent situations” rather than predict the future.
The work described in this paper is very well presented. The authors have provided separate boxes of text and accompanying reference lists for a summary of other work in automatic monitoring, definitions related to surveillance elements, and definitions of their model’s normality constraints and normality degrees. Their description of their system’s architecture is clear and well organized. The figures and tables that accompany the discussion of experiments they have conducted are also well done and reinforce the content of the paper. This paper should be of interest to individuals whose work involves intelligent agents or the detection of anomalous behavior.