Key points are not available for this paper at this time.
We describe the design and implementation of a system architecture and algorithms to provide a networked multi-view video surveillance system. The system has been developed to deliver reliable tracking of pedestrians and vehicles through an outdoor environment imaged by many CCTV cameras. Each video workstation runs computer vision algorithms that perform single-view multi-object tracking, generating packets of symbolic object track data that are transmitted across the local network to both a local view-integration server and a track and video database server which provides remote access to highly compressed camera video and ground-plane displays. The view-integration processing enables higher-accuracy 3D object location and occlusion reasoning when objects are simultaneously seen in more than one camera view-field, and track handover for objects that move between spatially adjacent views. (5 pages)
Tim Ellis (Wed,) studied this question.