Wireless Sensor Networks for Detection of IED Emplacement

Abstract : We are investigating the use of wireless nonimaging-sensor networks for the difficult problem of detection of suspicious behavior related to IED emplacement. Hardware for surveillance by nonimaging-sensor networks can cheaper than that for visual surveillance, can require much less computational effort by virtue of simpler algorithms, and can avoid problems of occlusion of view that occur with imaging sensors. We report on four parts of our investigation. First, we discuss some lessons we have learned from experiments with visual detection of deliberately staged suspicious behavior, which suggest that the magnitude of the acceleration vector of a tracked person is a key clue. Second, we describe experiments we conducted with tracking of moving objects in a simulated sensor network, showing that tracking is not always possible even with excellent sensor performance due to the ill-conditioned nature of the mathematical problems involved. Third, we report on experiments we did with tracking from acoustic data of explosions during a NATO test. Fourth, we report on experiments we did with people crossing a live sensor network. We conclude that nonimaging-sensor networks can detect a variety of suspicious behavior, but implementation needs to address a number of tricky problems.