【作者】 毕方明；

【导师】 张虹；

【作者基本信息】 中国矿业大学 ， 地图制图学与地理信息工程， 2010， 博士

【摘要】 随着空间信息需求的迅猛增长，空间信息处理已经步入集成化、网络化和产业化发展的新时期。现有空间信息处理的网络模型已经不能很好地适应空间信息的分布式处理和共享要求。随着对等网络技术的发展及其在分布式系统中广泛地应用，使用对等网络来解决大规模科学工程问题已成为一种趋势。为了对复杂的空间信息进行管理和维护，提高传输访问和处理效率，迫切需要有一种高效的空间信息处理网络模型。本文重点研究空间信息处理的P2P分层网络模型(SIPPLNM)和所涉及的空间信息分布、空间索引与查询、安全性技术。由于海量空间信息的处理中往往存在逻辑覆盖网络和物理网络不一致的问题，本文根据空间信息处理的特点，提出一种适用空间信息处理的P2P分层网络模型(SIPPLNM)。该模型分为两层：超级节点层和区域节点层，超级节点层是由负责相应区域的超级节点组成的，区域节点层由一个区域内的所有节点组成。根据空间信息的邻近性和非均匀分布特点，提出一种基于非均匀Hilbert曲线区域划分方法。通过非均匀Hilbert曲线保持空间对象之间的邻近性，提高空间信息划分区域的均匀性。区域内采用hash空间信息主题方式，实现空间信息在第二层节点的均衡分布。实验表明，本方法能够有效地克服现有空间信息区域划分和分布方法的不足，使其在SIPPLNM各节点中，均有良好的分布均衡性，并且能有效解决空间数据集在实际应用中不均匀分布的问题。分布式环境中的信任问题是当前网络研究的一个热点问题，而且在SIPPLNM下，因为存在空间关系，导致已有的信任管理模型无法直接使用。本文在基于主观逻辑理论和空间信息实体行为的基础上，提出了一种适合于SIPPLNM的分层信任模型，来处理空间信息实体之间的信任关系。该模型对超级节点间信任关系和域内信任关系分别采取不同的方法进行处理。实验和分析结果表明，这种信任模型能精确地评估空间信息实体之间的信任关系，从而能更加有效地解决对等环境中存在的安全问题。针对现有并行空间索引方法普遍存在的索引热点和访问瓶颈、以及空间索引结构一致性难以维护等主要缺点，提出了基于SIPPLNM的分布式、分层次的空间索引结构和空间查询算法。通过对索引结构的层次划分，将整个空间索引划分为多个子区域索引集合，从而将索引维护工作最大程度地限制在每个子区域内部，减小了分布式空间索引结构维护的复杂性；详细描述了SIPPLNM的索引操作算法和及其相关空间查询算法，并通过空间查询实验，验证了SIPPLNM索引结构的有效性及其快速的查询响应性能。应用实例分析表明，SIPPLNM能够有效处理非均匀空间信息，分层信任的机制也保证了SIPPLNM稳定运行。更多还原

【Abstract】 With the rapid development of spatial information, spatial information processing hasentered the new period characterized as being more integrated, industrialized and networkedclosely with network. The existing network model of spatial information processing alreadycan’t adapt to the request of distributed processing and sharing of spatial information.Meanwhile, with the development of peer-to-peer network and its comprehensive applicationin distributed system, people tend to solve large-scale scientific issues by using peer-to-peernetwork. In order to manage and maintain complex spatial information efficiently andimprove the efficiency of transmission and processing, there is an urgent need for a newnetwork model of spatial information processing. This paper deals with peer-to-peer layerednetwork model of spatial information processing(SIPPLNM) and involving technologies suchas spatial information distribution, spatial index and query, security.As some characteristics of spatial information processing and the mismatch between theoverlay and physical network in the management of the magnanimous spatial data, the paperproposes the peer-to-peer layered network model of spatial information processing(SIPPLNM), which consists of super-node layer and inner-district-node layer. Every district isin charged of one super node in the district, and the super-node layer consists of super nodes;the inner-district-node layer consists of all nodes of one district.According to adjacent relations and non-uniform distribution characteristics of spatialinformation, a geographic division method based on skewed Hilbert curve is presented. Byskewed Hilbert curve, the adjacent relations among objects are kept，and the balanceddistribution of spatial information among districts can be achieved. Inner the district, thebalanced distribution of spatial information in nodes of the second layer is achieved byhashing spatial information subject. Experiments show that this method can effectivelyovercome the drawbacks of existing geographic division and spatial information distributionmethods, and achieve a good storage balance in nodes of SIPPLNM. It is particularly suited tothe uneven distributed spatial data sets.Trust in distributed environment is a hot topic in network research, while current trustmanagement model can’t be used into SIPPLNM directly, for there are spatial relations inspatial information processing. Corresponding to SIPPLNM, the paper develops a layer trustmodel based on subject logic theory and spatial information entity behavior, which describestrust relationships between spatial information entities. It discriminates the trust relationshipsbetween spatial information entities in the same autonomous district and super nodes indifferent districts, and handles the two different trust relationships with different approaches. By simulated experiments, it can evaluate the trust relationships between spatial informationentities precisely, thus can resolve security problems existing in P2P environment effectively.After analyzing the inherent problems in general distributed spatial index methods suchas index hotspots, access bottlenecks and the difficulty to maintain consistency of distributedspatial index structure, a distributed layer spatial indexing structure and query algorithmsbased on SIPPLNM are presented. By layer indexing structure, the entire spatial index will bedivided into a sets of district index, which helps to restrict the works of index maintaining ineach district itself as much as possible and can reduces the complexity of maintainingdistributed spatial index structure; the detail of the SIPPLNM indexing algorithms as well asrelated query algorithms are illustrated in this paper. Spatial query experiments also verify thevalidity of the SIPPLNM structure and its rapid query response performance.By analyzing practical application, SIPPLNM can process non-uniform spatialinformation efficiently and run reliably after introducing layer trust model.更多还原