P2P-based geometric computation method for extracting ROI from ubiquitous video cameras

This paper proposes a method for extracting the ROI using geometric computations, to obtain information of commonly-interested objects by multiple users. The authors use P2P Delaunay Network to share the regional information shot by cameras with peers, and extract the most intersecting area of camera ranges as ROI. In P2P Delaunay Network, nodes and edges refer to as computers and their connections. This proposed method can construct P2P Delaunay Network using locations of mobile devices as GPS phones or PDAs equipped with a camera, and process computations distributively with these devices. Thereby, a P2P Delaunay network can be constituted, without using the computer resource of other settlements. The method first partitions the plane and assigns the partitions to nodes according to their locations using P2P Delaunay Network. Then, each camera propagates its camera range information to nodes, and each node thus processes computation distributively. The node which has received multiple camera range information computes the intersecting area within its Voronoi region. Finally, the ROI can be determined, by computing the most intersecting area among all nodes. The authors examine the proposed method by evaluating the frequency of region partitioning and data transfer for each node through numerical simulation.