【作者】 赵婵婵；

【导师】 冯秀芳；

【作者基本信息】 太原理工大学 ， 计算机应用技术， 2007， 硕士

【摘要】 无线传感器网络(Wireless Sensor Networks，简称WSN)集成了传感器技术、嵌入式计算技术、无线通信技术和分布式信息处理技术。它是一种全新的计算模式，是继因特网之后将对21世纪人类生活方式产生重大影响的IT热点技术。因特网改变的是人与人之间交流、沟通的方式，而无线传感器网络将逻辑上的信息世界与物理世界融合在一起，改变的是人与自然交互的方式。自组织特征映射网络(Self-Organizing Feature Map，简称SOM网络)是由芬兰赫尔辛基神经网络专家Kohonen于1982年首次提出的。它是竞争型人工神经网络模型，也称为Kohonen网络。从本质上说，是希望解决有关外界信息如何在人脑中自组织地形成概念的问题。自组织映射网络是一种具有聚类功能的神经网络，可以作为状态估计的一种新模型。信息融合能将来自多个传感器节点的信息加以联合、相关和组合，剔除冗余信息，获得互补信息，并对传送信息的重要程度进行适时的完整评价。信息融合技术应用在无线传感器网络中，可以在汇聚信息的过程中减少信息传输量，提高信息的精度、可信度以及网络收集信息的整体效率。本文从信息融合的本质出发，系统化地研究了基于SOM的信息融合技术。对无线传感器网络中的信息融合理论与技术进行了深入的剖析，主要工作如下：(1)总结了无线传感器网络独有的特点，讨论了关于无线传感器网络的一些研究热点问题和无线传感器网络协议体系。(2)系统地研究了信息融合理论，从定义、体系结构、分类、算法实现、应用领域等方面详尽地进行了阐述和分析。真正从用户角度出发考虑信息融合需求，将融合种类划分为三类：面向节点、面向区域、面向网络。接着引入软件工程中类的思想，把用于不同类别的融合方法和信息类型封装，可实现不局限于具体背景的应用。然后给出了无线传感器网络中基于UML(统一建模语言)的信息融合框架。(3)提出了无线传感器网络中基于SOM的信息融合系统原型，采用MATLAB实现了该信息融合系统。系统包括读取数据模块，融合预处理模块，信息融合模块，查看历史信息模块。实验证明该信息融合系统在煤矿环境安全监测方面有很好的效果。另外，系统采用文字与图形表示结合使用的方法，使融合结果更形象直观。综上，无线传感器网络信息融合技术用于处理海量信息，减少信息冗余以及信息传送量，降低了能量消耗，延长了网络使用寿命。基于SOM的无线传感器网络信息融合系统可同时处理网络中同质和异质信息，估计网络所监测环境的整体状态，并能给予智能决策上的辅助，这对于环境监测应用方面有着极其重要的意义。更多还原

【Abstract】 Wireless Sensor Networks (WSN) is integrated by the sensor technology, embedded computing technology, wireless communicating technology and distributed information processing technology. It is a novel calculation pattern which will produce the significant effect on 21 centuries human being life-style after Internet. Internet changes the interpersonal mode of communication and intercourse, but WSN fuses the logical information world and physics world, and then will change the interacting between human and nature.Self-Organizing Feature Map (SOM) is a competition artificial neural network (ANN) model. Neural networks expert Kohonen brought forward it first in Finland Helsinki in 1982, and it also be called the Kohonen network. Essentially, it makes environment information form a self-organizing concept in the human brain. SOM is a type of ANN with clustering function, and it provides a new idea for us to evaluate the state.The information fusion gets information which comes from many sensor nodes and information source unities. It uses the relevaince principle and combination, thus gains complementary information without redundancy. It can estimate timely condition and important degree of transferring information. Information fusion technology of wireless sensor networks can cut down information transmission amounts and improve information accuracy, confidence and the overall efficiency of collecting the information.Starting off the nature of information fusion, this thesis uses the systematization method to research information fusion, analyzing the theory and the technology of information fusion. Our main work is as follows:(1) summarizing the particular characteristics, a few of hotspot problems and protocol architecture in wireless sensor networks.(2) systematica lly study information fusion theory including definition, system structure, classification, algorithm, applying field. From user’s point of view, the necessary information can be undoubtedly divided into three types: node oriented, area oriented, network oriented. Next we import class thought, encapsulate the different fusion methods and information types, and then may realize the application without constrains. Therefore we give an information fusion framework based on UML in WSN.(3) bring forward information fusion system based on SOM. Simulation modules include reading information module, fusion pretreatment module, information fusion module, result display module and retrospect module. The simulated experimental result has confirmed validity of this method. Combining letter with graphics makes the fusion result more intuitionistic.In a word, SOM has the formidable ability in processing magnanimous information. Therefore, information fusion technology based on SOM in WSN can obtain more exact and general information of environment to assist intelligence decision-making. It is significance of environment monitoring.更多还原