Online Dynamic Event Region Detection Using Distributed Sensor Networks

Event region detection refers to the process of detecting regions with distinguishable characteristics in an environment, and it can find a broad range of applications from environmental monitoring to system health management. The problem of online dynamic event region detection is studied here. The spatiotemporal relationship of the evolving event regions is assumed and modeled by dynamic Markov random fields. Observations are collected from a network of sensors distributed in the field. To provide detection results at each time step, a distributed event region tracking algorithm is proposed. The system dynamics and information collected from neighbors are used to predict the underlying hypothesis at each sensor node and its local observation is used for update. Mean field approximation is adopted in the algorithm for tractability. The performance of the proposed algorithm is analyzed both theoretically and through simulations. By comparing with static event region detection algorithms and a centralized algorithm (with certain approximation), we demonstrate the effectiveness and efficiency of the proposed algorithm, especially its robustness in low signal-to-noise ratio (SNR) situations.