This paper presents the framework of a novel approach to combine multi-modal sensor information from a network of distributed smart nodes possessing audio and video modalities for gaining valuable supplementary information compared to traditional video-based surveillance systems or even just CCTV systems. The nodes are equipped with wireless communication modules allowing high-range, medium-bandwidth, and secure communication in a "cable-replacement" fashion. Some of the nodes also possess wired Ethernet connections to allow the transport of video streams to the supervisor. In a stepwise procedure the nodes first discover neighbor nodes by means of wireless connection strength and then try to find overlaps on the semantic level by utilization of the loopy belief propagation algorithm. This procedure ensures scalability over the network and allows establishing a global view of behavior in some region shared by several nodes. The processing architecture including the physical sensor nodes is called SENSE (smart embedded network of sensing entities) [1, 2].
展开▼