On accurate and efficient statistical counting in sensor-based surveillance systems

Sensor networks have been used in many surveillance, providing statistical information about monitored. Accurate counting information (e.g., the distribution the total number of targets) is often important for making. As a complementary solution to doublecounting communication, this paper presents the first that deals with double-counting in sensingfor wireless networks. The probability mass function (pmf) of counts is derived first. This, however, is shown to be prohibitive when a network becomes large. partitioning algorithm is then designed to significantly reduce complexity with a certain loss in counting. Finally, two methods are proposed to compensate the loss. To evaluate the design, we compare derived probability mass function with ground truth obtained exhaustive enumeration in small-scale networks. large-scale networks, where pmf ground truth is available, we compare the expected count with true target. We demonstrate that accurate counting within 1 rv 3% relative error can be achieved wit orders ofmagnitude in computation, compared with an exhaustive based approach