Study Of Efficient Event Boundary Detection Algorithm In Wireless Sensor Networks

Boundary detection is one of the most important research contents in event monitoringarea of wireless sensor networks. In process of event monitoring, when event that peopleinterested in happened, people are most concerned about its coverage and the distributionsituation of event boundary. Domestic and foreign scholars have done a lot research on thisissue. As the nodes of wireless sensor networks have the characteristics of vulnerability, limitin resource and energy and influence by environment etc., it can easily lead to generating offault sensor nodes whose error data can highly reduce the performance of event boundarydetection. While current boundary detection algorithms also have problems of low detectingaccuracy rate, high false positives rate and uncontrollable thickness of event boundary whichcause them being limited in practical application. To solve the problems above, this paperstudies high-efficiency event boundary detection solution and carries out some innovationwork as follows:Firstly, this paper proposes a reputation model based fault detection algorithm (RMFD).This algorithm uses the thought of existing spatial correlation and introduces a reputationmodel with punishment mechanism to quantify the faulty degree of nodes. It usesmulti-period dynamic adjustment to adjust node reputation and determine whether it is a faultnode according to its reputation value. In the process of Reputation adjustment, it usespunishment mechanism to increase the deduction amount of reputation of the nodes whoprovide unreliable readings repeatedly. The experiments’ results show that while ensuring ahigh recognition rate this algorithm can also significantly reduces false positives rate,detecting time delay and energy consumption.Secondly, this paper proposes a novel curve fitting based boundary detection algorithm(CFBD). This method uses the characteristic of event attribute value gradient-changing. Itfinds the sensor nodes whose data closest to the threshold value of happening event in theneighborhood, and then uses curve fitting technique on these nodes coordinates to get eventboundary curve equation, and finally every node can be visually determined as a boundarynode or not through the distance from point to the curve. The experiments’ results show thatthis algorithm has high performance on recognition rate, false positive rate and controlling ofboundary thickness.Finally, this paper gets an efficient boundary detection algorithm by integrating both ofthe algorithms above. Use RMFD to exclude the fault nodes in the wireless sensor networks, and then use DFBD to carry out efficient boundary detection. Through simulationexperiments, results have both higher recognition rate and lower false positive on boundarynodes than that of CFBD.