Energy-Efficient Distributed Target Tracking in Wireless Video Sensor Networks

摘要

In order to prolong the lifetime of wireless video sensor networks (WVSNs), many methods for visual target tracking are proposed to reduce the energy cost of battery-powered video sensor nodes. Focus on reducing the amounts of image data for computation and communication, this paper presents an energy-efficient method for distributed target tracking in WVSNs. The major contributions of the proposed approach are as follows. First, each activated video node uses real-time background subtraction based on balanced com-pressive sensing adaptive gaussian mixture Model for target detection. Second, targets in different views are associated by homograph projection of the position and hue histogram. Third, a Sign of Innovation based consensus filter is proposed for distributed target tracking. Experimental results show the method can effectively reduce the energy cost and achieve satisfying performance in real scenes.