Networked video surveillance using multiple omnidirectional cameras

Abstract

Remote surveillance is widely utilized in banks, shops, offices, at home and so on. In most conventional remote surveillance systems, fixed or active cameras with a narrow field of view are generally used in order to acquire an image of the remote site. This paper proposes a new networked surveillance system. The proposed surveillance system, which uses multiple omnidirectional cameras and network, is based on a server/client model: the server computers, each of which is connected to an omnidirectional video camera, are placed in the surveillance area and the client computer is placed on the user side. The servers detect moving objects and estimate their directions from sensors. The client estimates object positions from their directions received from the servers and presents object-centered perspective images to the user. In experiments, the implemented proposed system can do those at real-time.