A PHD-filter-based multitarget tracking algorithm for sensor networks

Abstract

Because of their applications potentials, sensor networks have attracted much attention in recent years. The problem addressed in this paper is multitarget tracking in sensor networks. In order to strike a balance of tradeoff between accuracy and energy consumption in tracking time-varying number of targets in sensor networks, we propose an energy-efficient multitarget tracking algorithm based on the probability hypothesis density (PHD) filter. We first analyze the PHD-filter-based hierarchical fusion architecture within a two-level fusion scheme running respectively at the cluster heads and base station of the network. Using a prediction-based approach, a dynamic sensor selection scheme is further examined. Simulation results demonstrate the capability and effectiveness of the proposed algorithm in terms of energy efficiency and tracking accuracy. It shows that our proposed algorithm is an attractive energy-efficient approach to track time-varying number of targets in sensor networks. © 2013 Springer-Verlag Berlin Heidelberg.