Inference Framework for Smart Surveillance

Video surveillance is rapidly growing in the area of public safety, facilities surveillance, and traffic monitoring. In visual sensor network, the visual information of the monitored field from any arbitrary point is provided to user. In a centralized system, video streams from different cameras are sent to a control system where human operator monitors it. This paper proposes enhancement in previously used algorithm to accomplish scalable smart video surveillance. Previous work rely on an algorithm for work-tree formation, which is designed to reduce the redundant data processing by maximizing the overlapping index [1]. The new approach takes account of battery and packet-drop at each node. This new approach when implemented will solve the problem of overloading at the base node, and loss of information in case of low battery and packet-drop.