Online Event Detection Based on the Spatio-Temporal Analysis in the River Sensor Networks

Events detection with the spatio-temporal correlation is one of the most popular applications in the wireless sensor networks. In general, the existing approaches separate time and space data properties and cannot combine abnormal characteristics of the global network to make a unified spatial-temporal detection. In this paper, a decentralized algorithm based on probabilistic graphical models (PGMs) of spatial-temporal detection was proposed to detect abnormal event with spatio-temporal correlation. Firstly we utilize the connected dominating set (CDS) algorithm to select backbone nodes to avoid collecting a large amount of sensory data from all the sensor nodes. Then, adopting Markov chains to model the temporal dependency among the different sensor nodes, and Bayesian Network was applied to model the spatial dependency of sensors. Based on the analysis of the spatio-temporal data correlation, the wireless sensor network can predict the abnormal events occurrence. In the paper, the online event detection with spatio-temporal correlation is applied in the river sensor networks. The extensive experimental results demonstrated that the proposed algorithm can achieve better performance than the simple thresholds algorithm and Bayesian Network based algorithm in the terms of the detection precision, data transmission delay, scalability and response speed.