A System for Tracking and Recognizing Multiple People with Multiple Camera

Abstract

In this paper we present a robust real-time method for tracking and recognizing multiple people with multiple cameras. Our method uses both static and Pan-Tilt-Zoom (PTZ) cameras to provide visual attention. The PTZ camera system uses face recognition to register people in the scene and ``lock-on'' to those individuals. The static camera system provides a global view of the environment and is used to re-adjust the tracking of the system when the PTZ cameras lose their targets. The system works well even when people occlude one another. The underlying visual processes rely on color segmentation, movement tracking and shape information to locate target candidates. Color indexing and face recognition modules help register these candidates with the system.