【作者】 邵世维；

【导师】 傅仲良；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2011， 博士

【摘要】 随着地理信息系统(下文简称GIS)技术的不断发展和广泛应用,多源、多尺度、多时相空间数据的快速更新和集成应用也成为一种必然趋势,对GIS空间数据库更新与可持续发展的研究与探讨已经成为迫在眉睫的任务,矢量空间数据的匹配技术也已成为迫切需要解决的关键问题。由于同一地区的矢量空间数据往往被多个部门重复采集或同一部门不同时间采集,这些数据在几何位置、几何形状、拓扑结构、几何精度、属性详细程度、编码方案、语义表达以及实体空间关系等方面不尽相同,使得数据的合并与共享非常困难。为了得到精度较高、属性信息更丰富、地图范围更大的高质量的空间数据,就需要对不同的地图数据库之间进行实体匹配,建立实体之间的连接,并在此基础上进行矢量空间数据融合处理,解决数据之间的几何、拓扑和语义的不一致性问题。本文对面实体的几何特征描述方法进行了深入的研究,并探讨了面实体在多尺度下的特征变化及其匹配方法,研究了实体匹配技术在数据库更新中的应用,给出了合理可行的多尺度实体匹配方法和基于要素级数据更新方案。首先,本文针对面状矢量数据的几何特征及矢量要素类型的不同,设计了四种几何描述方法,分别为重心射线法、凸包法、正切空间法和离散几何矩法。重心射线法和凸包法可以对面实体进行粗匹配,具有算法简单,速度快的特点。为了精确地描述面实体几何特征,同时最大限度保持实体的原有信息,本文提出了基于正切空间的特征描述方法,将面实体通过正切空间的方式进行描述,通过形状相似度距离对匹配实体进行几何相似度计算,实验证明,利用正切空间的思想描述面实体,描述方法简单,存储计算量小,其描述函数对目标具有形状不变性,对复杂面实体具有匹配准确的特点。本文还提出了一种适用于矢量图形的形状不变几何矩描述方法,将传统的基于图像像素点的离散不变矩改进为由面实体轮廓构建的三角形矩累加,解决了不等间距离散矢量数据的带来的问题,提高了匹配的准确率。其次,考虑对跨尺度的一对一实体匹配,通过Douglas-Peucker算法对面实体建立层次模型和然后再用角度演算算法对形状分段后的图形进行简化,将简化后实体再进行实体匹配,提高了匹配的正确率。针对多尺度匹配中的一对多匹配,提出了一种基于外扩和内收缓冲区聚类方法的匹配模型,通过缓冲区合并算法提取多个面实体的外轮廓特征,再与源匹配实体进行几何匹配。此外,还研究了基于面与面之间的基本空间拓扑关系模型,并通过建立三级混合索引的方式对空间索引进行优化,通过固定格网索引对候选匹配集进行快速准确查找,避免了叠加分析等计算操作,并最大限度的缩小了匹配集的范围。最后,设计了一整套基于实体匹配的要素级数据库更新处理流程,通过基于权重及规则的相似性计算模型,解决了多种要素类同时进行匹配等复杂情况下的面实体匹配问题；通过实体匹配后的相似度结果来判断实体是否需要更新,并将已更新实体放入历史数据库中,方便要素的历史管理。并通过地理对象的建立,来解决多尺度下的联动更新问题。更多还原

【Abstract】 With the development and wide application Geographic Information System (hereinafter referred to as GIS) technology, integration and instant (quickly) updating of multi-source, multi-scale, multi-temporal spatial data, has become an inevitable trend quickly update and integration of applications. Thus quickly updating and sustainable development of GIS spatial database has become an urgent task, and the vector spatial data matching technology is the key issue to be addressed. Since Vector Spatial Data of the same region are usually repeatedly collected by a number of departments or the same departments but at different times, the data varies in the geometry position, geometry shape, topology structure, geometric accuracy, and level of attribute detail, coding schemes, semantic expression and the relationship between entity spaces. This makes the consolidation and sharing of the data become very difficult. In order to obtain spatial data of higher accuracy, richer attribute information and wider coverage of mapping with high quality, the procedure of entity matching is very necessary between different map databases to establish the relationship between entities. And on this basis, fusion of vector spatial data can be handled to solve geometric, topological, semantic inconsistencies between the data.This thesis makes thorough research on methods of entities matching, and explores changes of the surface entities features in a multi-scale and its matching methods. In Addition, this thesis illustrates applications of entities matching in the database updating. This thesis also proposes reasonable and feasible methods of multi-scale entities matching and schemes of element-level data updating.Firstly, according to differences of geometric features of surface vector data and type of vector elements, three geometric matching methods are designed, that is, orientation coding method, tangent space method and the discrete geometric moments. Orientation coding method applies for matching of simple and regular surface entities at small and medium scale (such as 1:50000 buildings, streets, etc.), with a high processing speed. In order to identify the geometric characteristics(features) of surface entities, while to maintain more information about the original attribute of the entities as possible, tangent space method(matching method on the tangent space) is proposed. This method expresses (projects) shapes of surface entities in the tangent space, and calculates geometric similarity with adaptive similarity of shapes. Experimental results show that describing surface entities in tangent plane, feature extraction method is simple with little memory, and the coding has the characteristics of shape invariance. Additionally, this methods obtains accurate matching when applied to complex surface entities at small and medium scales. When there is variance in scales (When there is a certain span of time scales), this thesis introduces improved shape invariant geometric moments to the process of matching. By adjusting the traditional discrete HU moment invariants based on image pixels to moments constructed by triangles witch are established by outline of surface entities, matching accuracy has been improved.Secondly, considering the span of a certain scale of one to one entity matching, hierarchical model of surface entities can be built with the Douglas-Peucker algorithm and angle calculation algorithm of entities, first extract the corresponding physical characteristics, and then a physical match and improve the matching accuracy. For one to many matches in multi-scale matching, this thesis puts forward a matching model based on clustering through expanding and contracting the buffer zone. This method extracts features of mathematical morphology operators and tangent space extract multiple surfaces merging algorithm outline entities attributes, and on this basis, geometric matching with the source matching entities will be performed.In addition, this thesis also researches on the basic topological relationship between surface entities, and in order to address the problem of unable to distinguish entities in geometric matching, a match method based on topological relationship of the overall environment is proposed. Topological matching happens according to the hierarchical relationship between the matching entities with the environment around. In the case of semantic multi-scale matching in multi-scale matching, semantic similarity is calculated through the establishment of semantic encoding and hierarchical properties of ontology structure and similarity of attributes, such as numeric and text strings can be calculated through the semantic relationship between entities.Finally, this thesis designs the whole processes of updating elemental database based on entity matching, and with index-based grid searching method, efficiency of searching candidate matching set during surface entities matching and efficiency of spatial analysis have been greatly improved. By adopting similarity calculation model based on weights and rule-based, surface entities matching problems in the complex situations with diverse elements are solved. Additionally, the similarity gained during the entities matching can be used to decide whether to update the database or not, and updated entities are settled down in historical database which facilitate elements management.更多还原