Efficient target tracking through binary-detection in sparsely deployed WSN

The problem of tracking moving objects with help of wireless sensor network (WSN) has been studied in past. Most of the solutions rely on the use of specialized and expensive sensors, and on dense deployment of sensors. These techniques are infeasible for applications in low budget domains. In this paper, we propose two novel techniques to track targets using binary sensing that does not need overlapping sensing regions. These techniques can track a target, and estimate the distance it has traversed on the basis of the time that target spends in the vicinity of sensors. In the first technique, the path traced by a moving target is approximated by tangent estimations to three circles, each representing range of a sensor. It allows us to convert the original problem into a semi definite program. The other tracking scheme identifies a band of small width where the target is guaranteed to lie. The band is first approximated using the distance travelled after coming out of vicinity of one sensor and before entering the vicinity of second sensor. This band is then reduced using the distance travelled inside the vicinity of the sensors. We simulated the two methods in NS2 and evaluate both the methods.