Research On Data Fusion Technology Of Wireless Sensor Network

Since the 21st century, wireless sensor network, pushed up by the development of embedded technology, wireless communication technology and distributed technology, is increasingly mature, and becoming one of hot topics in the field of wireless networks. The unique advantages of wireless sensor networks in military, medical, industrial, agricultural and other aspects of the application make it get more and more extensive attention. Wireless sensor network is a multi-hop, self-organized wireless network constituted by a large number of sensor nodes which deployed in the target area to perform specific tasks. But unlike the traditional wireless networks, the life of the network, communication bandwidth, computation ability of wireless sensor networks are very limited, becoming the main reasons of obstacle to the further large-scale application.The idea of data fusion is combining the large amounts of information from disparate data sources into a small amount of result which is closer to user needs. It can eliminate the redundancy of information, reduce the network traffic and transmission collision, thereby contributing to break the energy constraint of wireless sensor network, prolong the network life cycle, while improve the efficiency and accuracy of data collection. So the research of data fusion technology in wireless sensor network is very valuable.This thesis based on summarizing the current situation of wireless sensor network and data fusion, puts forward a set of wireless sensor network data fusion scheme. First from the angle of security fusion, the thesis proposes a reputation model based on similarity measure of trust value in wireless sensor network—RMSMTV. This model can effectively solve the existing problems in the reputation models, such as malicious evaluation of nodes, too much subjective factors in evaluation process. So RMSMTV provides a credible environment for the implementation of data fusion. Based on cluster-tree network structure, the fusion scheme mainly divides into network setting stage and steady- working stage. In network setting stage, selecting cluster head node is connected with the trust values of nodes and their energy, while establishing a routing tree among cluster head nodes. In the steady-working stage, member nodes in cluster send data to the cluster head node within the time slot allocated, cluster head node sends the results of data fusion forwarded to its parent node along the routing tree, so until to root node. The root node transmits eventual result to base station. In order to avoid the interference of external noise and other undesirable factors, when data fusion is executing in cluster head node, the fusion weighted proportion for each data is the ratio of mean and standard deviation of consistency measure, which could effectively eliminate the "outliers" in data fusion, reduce errors and ensure the accuracy of fusion result.