【作者】 董辉；

【导师】 孙优贤； 卢建刚；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2007， 博士

【摘要】 无线传感器网络(WSN)是由大量无处不在的、具有无线通信与计算能力的微小传感器节点构成的自组织分布式网络系统，是能根据环境自主完成指定任务的智能系统；是继Internet网之后随着传感器技术、微机电技术、无线通信技术和分布式信息处理技术发展而发展起来的一种新兴信息获取技术，将改变人类认识自然及与自然界交互的方式。传感器网络是计算机科学技术的一个新的研究领域，具有十分广阔的应用前景，引起了学术界和工业界的高度重视。无线传感器网络是以信息处理为中心的系统。无线传感器网络主要用于从环境中收集监测对象信息，对其进行处理，并传送到服务中心或用户。任何基于传感器网络的应用系统都离不开信息的管理和处理技术。信息处理技术主要包括数据的压缩、融合、存储、查询、分析、挖掘、理解以及基于感知数据决策、行为的理论和技术。无线传感器网络信息处理的目的是在资源受限和能量最优的条件下管理传感器网络监测区域的感知信息，使用户如同使用通常的数据库管理系统和数据处理系统一样自如地在传感器网络上进行感知数据的管理和处理，而无需关心传感器网络实现的细节。但由于无线传感器网络节点数量众多，分布密集，拓扑结构复杂多变，每个节点的资源极其有限(包括计算、存储、通信能力)，给无线传感器网络的信息处理带来了严峻挑战，为此，必须采取简单、高效、可扩展性好的处理策略。到目前为止，感知数据管理和处理技术的研究还不多，还有大量的问题需要解决。本文对无线传感器网络中的感知数据(信息)管理和处理等相关热点问题进行了研究，为进一步的应用和研究打下基础。本文所作的主要工作如下：1)提出了一种新型的基于简化提升格式的分布式小波数据压缩算法。这种算法计算量很小，能有效消除无线传感器网络中各个节点间的信息冗余，大大节省了无线通信所消耗的能量，延长了网络寿命。2)针对无线传感器网络中环境监测时的图像采集，本文提出了一种分布式小波图像压缩算法。利用监测节点所采集的图像通常具有的变化少、相关性强等特点，引入了简单的变化检测算法和位置估计／补偿算法检测并压缩图像的活跃区域，消除各节点所采集的图像之间的信息冗余，从而节省通信能耗。3)利用小波分解系数的多分辨分析和由粗及精等特点，提出了一种基于小波框架的新型分布式存储和查询方法。它使所有节点所采集的数据经小波压缩后平均分布于各个节点之中，构成小波系数空间存储结构树。提高了存储效率，并支持时空两个方向的快速查询。4)针对无线链路通信质量不可靠的情况，研究了无线传感器网络中基于无线链路的信息决策融合的问题，提出了一种新的最优似然比(LR)规则。考虑到传感器网络的资源受限，对信噪比较高或较低的情况下又提出了两种简化的次优融合规则，最后讨论了同类传感器模型下的融合情况。5)通过对传统Apriori算法的改进，提出了一种节点关联规则挖掘方法来发现大量节点之间的有趣关联或相关联系，以此消除节点之间的信息冗余。这种方法可以帮助用户对数据进行有效的融合、分类、查询、分析、理解和决策；减少信息处理中通信和计算所消耗的能量。最后总结全文，指出有待于进一步研究的课题和今后的重点工作。更多还原

【Abstract】 Wireless Sensor Networks (WSN) is a self-organized distributed intelligent system comprising low-cost, low-power, multifunctional sensor nodes that are small in size and capable of wireless communication in short distances. WSN, which integrates the technologies of sensor micro electro mechanical, wireless communications and distributed computation, is a novel mode of computing and a hotspot of information technology after internet. It will change the way that human recognize and communicate with the physical nature. Sensor network is a new research area of computer science and technology and has a wide application in the future. Both academia and industries are very interested in it.Wireless Sensor Networks is an information-centric system. The basic role of the WSN is to collect information of monitored objects in the environment, and then to process and transmit them to the sink node. Any application of wireless sensor networks is related to data management and information processing. Information processing in WSN includes data compression algorithms, distributed database storage and query strategies, in-network data fusion or aggregation and so on. The main goal of information processing in WSN is to manage the sensing data collected from the environment with limited resources and energy save. So that, the users can manages and process the sensing data as in the same convenient way as processing on the normal database without consider the detail of WSN.However, the characteristics of WSN bring great challenges to the information processing in WSN, such as the ultra large number of sensor nodes, dense deployment, frequently changing topology structure, the limited resources, including computation, storage and communication capability, and so on. All these require the information processing algorithms to be energy-efficient, scalable well and robust.This thesis gives an overview of the former work on the information processing approaches in WSN, and studies the related issues of information processing of WSN. The major contribution of this dissertation is stated as follows:1) This thesis proposed a novel distributed wavelet compression algorithm basing on simple lifting factorization for reducing energy consumption in wireless sensor network. This low complexity algorithm canexploit the inherent correlations that exist in or between sensor readings and significantly reduce transmission costs, and then prolong the network lifetime.2) Focusing on the application requirements in environment surveillance, a novel distributed 3D wavelet image compression algorithm was developed. Since surveillance image sequences are often characterized by low motion and high correlation in WSN, Low-complexity Change Detection Algorithm (LCDA) as well as Position Estimation and Compensation Algorithm (PE&CA) was designed to reduce the energy consumption by the way detect and compress interested regions, remove the correlation of the image collected by each sensor.3) Based on the multi-resolution analysis of wavelet, this thesis proposed a novel distributed storage/query method, which distributed data to every sensor node, and constructed a wavelet spatial architecture tree. This approach can store data efficiently and support highly efficient spatio-temporal query.4) The quality of link communication has a significant impact on upper-layer application in wireless sensor networks. We study the problem of fusing decisions transmitted over fading channels in a wireless sensor network. We propose a new likelihood ratio (LR)-based fusion rule which requires only the knowledge of channel statistics. Two suboptimum alternatives are presented in the low or high SNR channel to reduce the energy consumption. Under the identical local sensor and the same channel SNR (singnal noise rate) assumption, we study the fusion statistic in the last.5) Based on modified Apriori algorithm, this thesis proposes a novel node association rule mining algorithm for exploiting the inherent correlations between sensor readings. This approach can help users to manage data efficiently during aggregation, classification, prediction, query, understanding and decision-making, and then reduce computation and communication energy in information processing effectively.In the last chapter, we conclude the dissertation and outlines future research.更多还原