A modular area surveillance system

This paper describes the implementation of a modular, decentralized surveillance system which incorporates a range of sensors including TV cameras, ultrasonic and infra-red devices, and optical barriers. The key features of the approach are fault tolerance, modularity, and scalability, which are achieved through the use of a Decentralized Kalman Filter (DKF) as the main data association and tracking technique, and Dempster-Schafer evidential reasoning as the basis for combining estimates of target identity from the various sensors. The surveillance system is able to track multiple objects from multiple sensors, and can provide estimates of target identity fused over time. Results are communicated to the user through a graphical interface implemented under X Windows, which supports the designation of ''exclusion zones'' where the user can specify conditions for the triggering of an alarm. The paper is organized as follows. First, the authors discuss briefly the implementation of the surveillance system, using TV cameras as the main sensing devices and transputers for processing. Then the authors present the results of a limited experimental evaluation of the performance of the system. Finally, some conclusions are drawn and suggestions made for possible further work