Efficient Shape Recognition of Dynamic Event Regions Using Wireless Sensor Networks

Abstract

In this paper, we consider the problem of recognizing the shape of dynamic event regions in wireless sensor networks (WSNs). A key idea of our proposed algorithm is to use the notion of distance fielddefined by the hop count from the boundary of event regions. By constructing such a field, we can easily identify several critical points in each event region (e.g., local maximum and saddle point) which could be effectively used to characterize the shape and the movement of such event regions. The communication cost required for the shape recognition of dynamic event regions significantly decreases compared with a naive centralized scheme by selectively allowing those critical points to send a certification message to the boundary of the event region and a notification message to the data aggregation points. The performance of the proposed scheme is evaluated by simulations. The simulation results indicate that: 1) the number of messages transmissions during a shape recognition significantly decreases compared with a naive centralized scheme; 2) the accuracy of shape recognition depends on the density of the underlying WSN, while it is robust against the lack of sensors in a particular region in the field, and 3) the proposed event tracking scheme correctly recognizes the movement of an event region with small number of message transmissions compared to a centralized scheme.