【作者】 李哲涛；

【导师】 李仁发；

【作者基本信息】 湖南大学 ， 计算机应用技术， 2010， 博士

【摘要】 随着制造工业的日益进步、传感器网络广泛深入的应用和用户对网络具有多媒体功能的迫切需求,多媒体无线传感器网络受到前所未有的关注。它具有采集信息丰富、数据通信量巨大、各处理单元能耗相当等个性特点。正是这些特点,使得传统标量无线传感器网络的研究成果难以直接移植到多媒体无线传感器网络之中。在多媒体无线传感器网络的各分支研究中,时间同步、路由协议和图像压缩都属于它的关键技术。其中,时间同步是任何分布式处理系统的支撑技术。精确的时间同步是实现传感器网络协同工作、定位、移动目标跟踪等技术的基础。地理路由具有分布式决策、复杂度低、可扩展性强和能量相对高效等优点。因此,它被认为是最有希望和可能解决多媒体传感器网络路由问题的技术之一。多媒体节点强大的采集能力,导致数据量骤增,对节点和网络存储、传输带来极大挑战。图像压缩是节省存储空间、减少数据通信量和节约带宽的重要技术之一。它间接影响网络的应用范围、生命周期和工作效率。本文主要围绕时间同步、地理路由和图像压缩三个方面开展研究。多媒体无线传感器网络对时间同步算法在能耗方面要求更高效、在同步精度方面要求更精确。目前针对时间同步和测距的研究基本都是独立开展,即使有联合研究,也是在时间同步的基础上,讨论测距或定位问题。事实上,时间同步和节点测距都需要消息交换。基于这一共性,本文提出了一种基于测距的时间同步算法,协同实现时间同步和节点测距,它具有能耗低和同步精度高等特点。首先,逆向分析时间同步和节点测距之间的关系。接着,提出基于测距的时间同步算法原理和算法步骤。在同步方面,该算法可实现全网同步以及外同步：在测距方面,该算法可实现相对测距和绝对测距。然后,对算法在通信量和收敛时间方面进行了理论分析。最后,从消息交换量、同步精度、鲁棒性等几个方面验证了该算法较TPSN (timing-sync protocol for sensor networks)和RBS (reference broadcast synchronization)有一定优势。现有的地理路由算法主要集中于研究如何使单包传输时的路由性能最优。在多媒体无线传感器网络中,一次通信的数据量较大,如何联合信源数据特征进行路由是一种新的路由方式。通过分析多媒体无线传感器网络中采集的数据流特征,本文提出了渐进优化的地理路由算法。其设计的创新之处是：不再以路由一个包的路由性能作为设计目标,而是将一次会话中所有交互数据作为整体,使得平均每包传输的路由性能最优。通过利用先行包在路由方面的参考信息以优化后续包的路由,实现边路由边优化。该算法具有收敛于最短路径、可嵌入其它地理路由算法中的优点。本文以将渐进优化思想应用于GPSR(greedy perimeter stateless routing)为例,验证了该路由算法能够有效地减少GPSR协议在周边转发模式中由于绕道产生的跳数,平均路由跳数性能有明显改善。在图像压缩方面,首先对现有的基于小波变换的嵌入式编码算法进行了总结,分析了各算法的优缺点。然后,在深入分析经典压缩算法SPIHT(Set Partitioning In Hierarchical Trees)的过程中,发现该算法在D型集合分裂时没有遵循先整体后个体的思路,导致过早分裂O型集合。另外,该算法中各集合状态间还存在一定的状态冗余。因此,本文提出改进的SPIHT嵌入式图像压缩算法,主要工作包括：(1)改进D型集合分裂时,实现先整体后个体的编码方式；(2)消除D/L/O集合间存在的状态冗余。通过标准图像测试,验证了改进后的算法在集合分裂次数、PSNR等方面有一定的改进。更多还原

【Abstract】 With the increasingly advancement of manufacturing industry, the further extensive application of sensor network and the pressing requirements for sensor network’s function of multimedia from users, multimedia wireless sensor networks (MWSN) have attracted unprecedented attention. It has some unique characteristics, such as containing rich information, enormous data communication, and equal energy-consumption of each unit and so on. Because of these special characteristics of MWSN, it was very difficult to transplant research results of traditional scalar wireless sensor networks (SWSN) to MWSN directly. In various branch research hotspots of multimedia wireless sensor networks, time synchronization, routing protocols and image compression are parts of its key technologies. Among these, clock synchronization is an important service in any distributed system, including SWSN and MWSN. Most application of sensor network requires accurate clock synchronization among sensor nodes which includes coordinated work between nodes, location and moving target tracking. In all routing protocols, geographic routing has been regarded as one of the promising routing methods for sensor networks due to its simplicity and scalability. Multimedia node has a powerful ability to capture information, which leads to a surge in the amount of data and brings a challenge to store and transmit it for nodes and network. Image compression is an important way to save storage space, reduce data traffic and save bandwidth. It indirectly affects the scope of network application, the lifecycle and efficiency of network. This thesis focuses on the research of time synchronization, geographic routing, and image compression.Compared to SWSN, MWSN require high efficient in energy consumption and high synchronization precision. Existing researches on time synchronization and ranging are independent. Even if there was joint research, it was synchronization-based ranging and location. However, all time synchronization methods rely on some sorts of message exchange between nodes, the same to range estimation. Based on the common characteristic of time synchronization and range estimation, this thesis proposes a novel hybrid scheme:range-based time synchronization protocol, which has higher efficiency of energy consumption and smaller synchronization error. It can reach time synchronization and range estimation at the same time. The first step is to analyze the internal relationship between time and distance conversely. Secondly, range-based time synchronization protocol and its steps are introduced. As for time synchronization, it can achieve network-wide synchronization and external synchronization. As for range estimation, it can achieve relative ranging and absolute ranging. Finally, we do some simulations on the platform of NS2. The results show it outperforms timing-sync protocol for sensor networks (TPSN) and reference broadcast synchronization (RBS) in terms of the number of message exchanges and synchronization error.Existing geographic routing algorithms mainly focus on how to optimize the routing performance of transmitting single packet. In multimedia wireless sensor networks, data communication transmission is huge. It is a new way to combine source data characteristics for designing routing protocol. By analyzing the data flow characteristics of multimedia wireless sensor networks, this thesis presents asymptotically optimal geographical routing (AOGR) for MWSN. Its design goal is to optimize the average routing performance per packet. In AOGR, the reference of delivered packets to subsequent packets in routing will be exploited. By selecting some nodes as waypoints, the route will be controlled by the source node and the destination node plus these waypoints. It possesses desired properties:routing gradually converges to the optimal path; it can be embedded in existing geographical routing protocols. We have embedded AOGR to GPSR (greedy perimeter stateless routing) in this thesis. The average performance of AOGR is compared to GPSR and the benchmark shortest path algorithm. Simulation results show that on average the proposed algorithm can reduce as much as 50% of hops on the routes obtained by GPSR.As for image compression, the thesis summarized existing wavelet-based embedded coding algorithms and analyzed their advantages and disadvantages. In the advanced process of analyzing SPIHT (set partitioning in hierarchical trees), the author found that it did not follow the rules that "set first, element second" in D-set partitioning, which leading to prematurely split O-set. In addition, there is some state redundancy between different kinds of sets. This thesis presents an improved SPIHT embedded image compression algorithm and some advice of how to optimize SPIHT is also introduced. Main tasks are:(1) in coding procedure, the idea which set coding is prior to element coding is realized by estimating the importance of O set. (2) The redundancy of states between set and subset, node and its leaves are eliminated by predicting code. The experimental results indicate that the improved algorithm enhances the ratio of important coefficient in output bit stream and consequently obtains the PSNR which is superior to the SPIHT algorithm.更多还原