Abstract

This paper introduces a generic architecture for the fusion of perceptual processes and its application in real-time object tracking. In this architecture, the well known anchoring approach is, by integrating techniques from information fusion, extended to multi-modal anchoring so as to be applicable in a multi-process environment. The system architecture is designed to be applicable in a generic way, independent of specific application domains and of the characteristics of the underlying sensory processes. It is shown that, by combining multiple independent video-based detection methods, the generic multi-modal anchoring approach can be successfully employed for real-time person tracking in difficult environments