Third IEEE International Workshop on Performance Evaluation of Tracking and Surveillance ( PETS ’ 2002 )

In earlier work we introduced a generic, modular tracker architecture that combines the advantages of several simple and rapidly performing tracking algorithms. The adaptive choice of critical system parameters such as processing regions and resolution results in robustness to varying frame rates and computational constraints. In this paper, we describe the embedding of our tracker into a distributed infrastructure for visual surveillance applications via an event-based mechanism. The tracker generates applicationindependent events on the basis of generic incidents and target interactions detected in the video stream. These events can then be received and interpreted by applicationspecific clients. We report experimental results on the shopwindow datasets of PETS 2002.