Research On Event Detection Algorithm In Wireless Sensor Networks Based On Evidence Fusion

Event region detection and event boundary detection technology for wireless sensor networks are current and hot research. Because of instabilities of the wireless sensor networks environment and limitations of node resources, namely: low prices, carrying energy limited, etc., nodes are prone to detection error during operation. So the event detection method for wireless sensor networks must have the fault-tolerance function. At the same time, it must consider the problem of energy consumption. According to the above questions and needs, this paper's research has the vital significance.At first, this paper presents a fault-tolerant event region detection algorithms based on node's evidence. It utilizes the evidence theory to realize data uncertain reasoning. The algorithm is divided into two phases. In the first phases, it takes advantage of the temporal correlated information to obtain single node evidence and make preliminary judging. In the second phase, according to spatial correlated, it makes use of the weighted model of evidence theory to complete evidence fusion and event decisions, and adjusts node's weights so that weight lower nodes are ruled out and the communication between the nodes is reduced. The simulation results show that, failure node recognition rate of this article algorithm is still high, and false judgment rate is too low when the node's failure rate is higher.Meanwhile, this paper presents fault-tolerant event boundary detection algorithms based on tree of evidence. At first, it takes advantage of the temporal correlated information to reduce the impact of failure node on boundary detection. Then it can exclude some nodes by querying and collecting information of parent and child nodes. Finally it makes use of theory of evidence to process data of its neighbor nodes, and detects boundary nodes. Simulation results show that the paper's algorithm has higher accuracy and more obvious boundary contour when nodes are randomly distributed and uniformly distributed.