Research On Distributed Target Detection In Wireless Sensor Networks

Wireless Sensor Networks (WSNs) provides the most direct service to people in the nextgeneration network which has attracted much attention in the world. Distributed target detectiontechnology is the basic use of WSNs and it uses randomly deployed sensor nodes to detect thetarget signal,then local sensor nodes send the results to the fusion center after they make thecorresponding decisions. The fusion center makes the system decision by the specific decisionfusion rule according to the received judgments from the local sensor nodes.In this dissertation, we mainly focus on the distributed target detection of the local decisionalgorithm and system data fusion algorithm based on the studying of the wireless sensor networks.We analyse the Bayes and Neyman-Pearson. And then we put forward the multiply sampledetection, threshold adjusting algorithm and the hierarchical distributed target detection algorithmbased on clustering. The contributions of this dissertation are described as follows:(1) In the study of the distributed target detection algorithm based on single sample, wepropose the algorithm based on multiple samples. The algorithm abandons the disadvantages dueto single sample caused by randomness. It takes multiply sample of the observed signal intoconsideration and makes full use of them to improve the detection performance of the local sensornodes.(2) Based on the distributed target detection, this paper proposes an adaptive thresholdadjusting algorithm according to the characteristics of the complex background noise. For theinitial value of any given threshold, we can get the result closed to the theoretical value throughthe adaptive threshold adjusting algorithm without the formula of the threshold.(3) On the basis of the practical application in wireless sensor networks, we propose thehierarchical distributed target detection algorithm based on clustering that is suitable formonitoring a wide range of interests. The algorithm not only considers the attenuation of thetarget signal, but also takes consideration of the Rayleigh fading of the local sensor’s judgments.We deduce the fusion criterion of the cluster header from the original likelihood rate testingfunction.We verify the effectiveness and the practice of the fusion criterion algorithm based onMultiply Sample, At algorithm and LRT_R algorithm respectively by comparing them to the otheralgorithms in simulation experiments.