【作者】 翟永平；

【导师】 罗武胜；

【作者基本信息】 国防科学技术大学 ， 仪器科学与技术， 2007， 硕士

【摘要】 无线多媒体传感器网络(WMSNs)是在传统无线传感器网络(WSNs)的基础上发展起来的具有音频、视频、图像等多媒体信息感知功能的新型传感器网络。同只具有简单坏境数据采集功能的传统无线传感器网络相比,无线多媒体传感器网络能实现细粒度、精准信息的环境监测,在军事、民用、商业等领域具有非常广阔的应用前景。基于对无线多媒体传感器网络未来广阔应用前景的认识,尤其是考虑到其在构建未来海、陆、空、天一体化战场态势信息感知系统中的巨大应用价值,本论文在“863”国家高技术研究发展计划项目“XXX数据压缩技术”(项目编号:2006AA701121)的资助下,围绕无线多媒体传感器网络中的图像信息分布式压缩展开研究,现将论文主要研究结果与结论归纳如下:1.全面综述了无线多媒体传感器网络中的节点系统设计、MAC协议、路由协议、多媒体信息处理等几大关键技术的研究进展,深入分析了这几大关键技术研究面临的挑战及未来研究的方向。2.提出了一种基于双正交叠式变换(LBT)的适用于无线多媒体传感器网络的图像分布式处理算法。该算法利用LBT变换结构上可并行计算的特点,将图像的不同块的变换并行地在不同节点上实现,同时为了防止块效应,变换采用两级节点来增强图像块边缘之间的相关性,为了降低编码复杂性,对传统的零树编码算法进行了简化并在两级节点上“循环”交替实现。仿真结果表明,该方法能在计算、存储资源非常有限的节点上有效的实现图像编码和传输,和“集中”处理方式相比,可成倍地提高网络生命周期。在分析算法结构的过程中,我们提出了节点数据流的概念及分析方法,该方法在多节点多路数据的大型网络数据的分析处理中存在非常大的应用价值。3.研究了分布式信源编码这一新型的信息编码理论及两种实用的编码方案——应用伴随式的相关信源编码方法(DISCUS方法)和基于turbo码的分布式信源编码方法(Turbo-DSC),并重点对Turbo-DSC方法进行了仿真分析。仿真中通过对相关信源不同相关性的建模,分别分析了二进制对称信源(BSC)和高斯信源情况下信源间相关性强弱与解码误比特率间的关系,得出了一系列有益的结论。4.探讨了分布式信源实用编码方案在无线多媒体传感器网络中的应用。针对多节点图像压缩,通过简化模型,给出了一种相机阵列中的图像分布式压缩编码框架;针对单节点视频处理,给出了一种帧内编码、帧间解码的基于变换域的分布式视频处理框架。这两种框架都基于LBT变换和Turbo-DSC编码器,这主要是考虑到LBT变换的低复杂度及变换系数存在零树结构的优点和turbo码近信道容量的优点。更多还原

【Abstract】 Wireless Multimedia Sensor Networks (WMSNs), developed from traditional Wireless Sensor Networks(WSNs), is a novel sensor network integrating sensation functions of multimedia streaming (e.g. audio, video, image). Compared with traditional WSNs which can only measure simple scalar physical phenomena such as temperature, pressure, humidity, or location of objects, Wireless Multimedia Sensor Networks are able to ubiquitously retrieve multimedia content such as video ,audio streams and still images. Potential applications of Wireless Multimedia Sensor Networks span a wide spectrum from military to industrial, from commercial to environmental monitoring.Wireless Multimedia Sensor Networks have broad application foreground in military, especially in constructing the integrative information system in battle field .Supported by the 863 National High Technology Research and Development Program "xxx image data compression technology"(No.2006AA701121), this dissertation concentrates its attention on the distributed information processing in Wireless Multimedia Sensor Networks. The main contents and contributions of this dissertation are as follows:1. The state of art in the key techniques of WMSNs, including node systems, MAC protocols, routing protocols, multimedia signal processing, is surveyed, and open research issues are discussed in detail.2. A distributed image compression algorithm based on lapped biorthogonal transform (LBT) in the cluster-based Wireless Multimedia Sensor Networks is proposed. In this algorithm, the image is firstly partitioned into several tiles (blocks) that consist of eight rows, and then each block is sent to a node to do 1D LBT. In order to prevent blocking artifacts, this algorithm uses nodes in two levels to enhance the correlation of adjacent blocks. A simplified zerotree coding algorithms is used in "coding nodes" of two levels by turns to reduce the complexity of coding. Simulation results show that this distributed algorithm can prolong the lifetime of the WMSNs compared with "centralized" compression algorithm under a specific image quality requirement. The "node data stream" conception is proposed in order to analyse the distributed algorithm, this new conception and method will be useful in analyzing the data of sensor networks consist of thousands of sensor nodes on a large scale.3. The theory of Distributed Source Coding(DSC) is studied, and two predominant practical coding methods—Distributed source coding using syndromes (DISCUS) and turbo-DSC method based on punctured turbo codes are discussed. Then, this dissertation simulates and tests the turbo-DSC system using binary and Gaussian random sequences and finally gets a lot of useful results.4. The use of practical DSC coding methods in Wireless Multimedia Sensor Networks is discussed and two coding frameworks are proposed. One of the frameworks focuses on image distributed coding in camera array and the other on video coding in single sensor node. In the second framework, a scheme of translation-domain intra-frame encoder and inter-frame decoder is used. Thanks to low complexity and the zerotree structure in LBT coefficient and the excellent performance of trubo code, LBT and turbo-DSC method are used in the two proposed frameworks.更多还原