【作者】 翟仁健；

【导师】 武芳；

【作者基本信息】 解放军信息工程大学 ， 地图制图学与地理信息工程， 2011， 博士

【摘要】 随着空间信息获取技术的飞速发展,人类获取空间信息的能力较过去大大增强。国家各部门基于不同的应用目的和需求,生产了大量的空间数据,从而导致同一地区不同来源、不同类型、不同时相、不同精度、不同尺度的空间数据越来越多。如何对这些空间数据进行有效管理和综合应用是当前GIS发展中亟需解决的难点问题,其中尤其以空间数据的增量级联更新、集成与融合问题最为突出。上述问题有效解决的前提和关键是对不同版本空间数据中的同名要素进行自动识别,并建立它们之间的匹配关系,以此为基础,从而实现同名要素之间变化信息的自动发现与提取、语义信息和位置信息的融合等应用。实际上,同名要素匹配研究已经成为GIS诸多应用研究的一个必然需求和瓶颈问题,具有十分重要的理论意义和很高的应用价值。本文选取道路网与居民地这两类极具代表性的要素作为实验对象,基于全局一致性评价的思想,对多尺度矢量空间数据的匹配问题进行了研究,主要研究成果如下:(1)提出了基于全局一致性评价的多尺度矢量空间数据匹配的概念模型。该模型包含三个关键组成部分:数据源中要素之间空间结构特征的获取、匹配候选集的同步搜索和全局一致性评价。(2)提出了一种道路网空间结构特征的层次化表达方法:针对传统“结点—弧段”数据模型存在的不足,基于Gestal“t视觉连续性”原则并结合属性特征提取完整道路stroke,利用社会网络分析模型对道路的拓扑结构特征等级进行评价。提出基于道路等级分析构建层次道路网眼从而实现对道路网形态结构的层次剖分方法,通过对道路网眼的相邻关系和层次关系进行描述,建立了基于道路网眼的道路网空间结构特征的层次化表达模型。(3)提出了匹配候选集的同步搜索和整体评价的匹配策略:根据数据源中要素之间的空间结构关系,提取待匹配要素的邻域要素。将待匹配要素与邻域要素作为一个整体构成待匹配空间场景。对待匹配空间场景中的所有要素搜索匹配候选集,利用待匹配空间场景中要素之间的空间结构特征知识作为约束对匹配候选集中的要素进行筛选,从而大大减小匹配候选集的规模。根据待匹配要素的候选匹配要素的空间结构特征,将匹配候选集中的对象进行组合,构建了若干个候选匹配空间场景,为面向空间场景进行整体相似性评价奠定了重要基础。该方法在道路网匹配和居民地匹配中得到了实例验证,证明了它的正确性和有效性。(4)提出了基于全局一致性评价的多尺度矢量道路网匹配方法:选取道路路段作为匹配的基本单元,将道路路段划分为网眼型路段、树状型路段和独立路段三种,根据不同的路段类型充分利用相应的结构特征知识实施匹配;建立了面向匹配的道路网层次结构的数据组织方法;根据道路网的空间结构特征获取待匹配路段的邻域路段并构建一个待匹配局部道路网络,根据匹配候选集的同步搜索方法建立若干个候选局部道路网络,待匹配局部道路网络与每一个候选匹配局部道路网络之间建立了路段对路段、道路网络对道路网络相对应的相似性评价关系;提出了基于道路网眼的匹配判断方法;在匹配判断上,建立了以局部道路网络为整体的空间相似性评价模型,不仅考虑路段与候选匹配路段的个体相似性,同时考虑了待匹配局部道路网络与候选匹配局部道路网络的整体结构特征的相似性,从而保证了待匹配路段与邻域路段在匹配结果上的全局一致性,实验证明该方法具备较高的匹配率和匹配精度。(5)提出了基于全局一致性评价的多尺度矢量居民地匹配方法:以道路要素作为全局性约束,选取格式塔原则中的邻近性、相似性和共同方向三个格式塔因子作为局部约束,实现了居民地群结构特征的自动识别,从而获取了居民地之间的分布特征以及空间关系。将待匹配居民地与邻域居民地组成一个待匹配局部居民地群,将其作为一个整体利用“滑轮法”实现待匹配居民地与邻域居民地匹配候选集的同步搜索,得到若干个候选匹配局部居民地群。建立了以局部居民地群作为评价对象的空间相似性评价模型,该模型不仅考虑了居民地个体的重叠程度、大小、方向等特征,同时也较好地顾及了居民地之间结构特征的相似性。实验结果表明:该匹配方法能够有效克服位置误差带来的不利影响,不仅能够识别1:1的匹配关系,同时也能够较好地识别1:N和M:N匹配关系,具有较高的匹配精度。更多还原

【Abstract】 With the rapid improvement of acquisition technology of Geo-spatial information, the capabilities for acquisition of geo-spatial information have been greatly enhanced than before. A great deal of spatial data has been produced by various departments for different purposes and requirements of applications, resulting in the increasing data that represents the same region with different sources, different types, different phases, different precisions and different scales. So how to achieve effective management and comprehensive utilization of these spatial data are difficult issues that need to be urgently resolved in development of GIS, among these difficult issues, particularly incremental and propagative updating of spatial data, data integration and fusion. To automated recognize homologous features and build matching relationship between them from different versions of spatial data is the premise and crux of effective resolution of the above issues. Based on the above matching result, many applications, such as the discovery and extraction of change information, fusion of semantic and location information between homologous features, will be achieved. In fact, homologous feature matching has become an inevitable demand and bottleneck of numurous applications of geospatial data, which has very important theoretical and practice significance.Based on the idea of global consistency evaluation, the thesis selected the road networks and settlements that have typical features as the experimental subject to do research on multi-scale vector data matching. The main contribution of this paper is as follows:(1) A conceptual model of multi-scale vector spatial data matching based on global consistency evaluation is proposed. It consists of three critical components which are spatial structure characteristics recognition between features in datasets, synchronous search of potential matching candidates, global consistency evaluation, respectively.(2)It is proposed a hierarchical representation model of spatial structure characteristics of road networks. Aiming at the shortcomings of the traditional“Node-Arc”model, based on the principle of Gestalt“visual continuity”and consistency of properties, all road strokes have been extracted and characterized quantitatively the topological properties by social networks analysis model. It is also putted forward the construction of hierarchical road meshes based on analysis of road grade, thus being achieved hierarchical partition of morphology structure of road networks. Through describing adjacent and hierarchical relationships between road meshes, the hierarchical representation model of spatial structure characteristics of road networks based on road mesh has been established.(3)A matching strategy of synchronous search of potential matching candidates and global similarity evaluation is put forward. According to spatial structure relationship between features in data sources, neighborhood features of the selected feature to be matched are automated identified. The selected feature to be matched and its neighborhood features are treated as a whole to constitute a spatial scene to be matched. Potential matching candidates of all of these features in spatial scene are identified firstly, and then, incorrect potential matching candidates are excluded by making full advantage of the knowledge of spatial structure characteristics as a result the scale of potential matching candidates set decreases greatly. According to spatial structure characteristics of potential matching candidates of the selected feature to be matched, all of the objects in potential matching candidate sets will be combined to build several potential matching spatial scenes, laying an important foundation for global similarity evaluation of spatial scene. This method has been confirmed to be accuracy and effectiveness by doing experiment in road networks matching and settlement matching.(4) It is put forward a matching method for multi-scale vector road networks based on global consistency evaluation. Having selected the road section as the elementary matching unit, the thesis divided it into mesh-type section, tree-type section and independent section and made full use of the corresponding structure characteristics based on different section type in matching progress. A data organization method of hierarchical structure of road networks for data matching is proposed. The adjacent sections of selected road section to be matched are recognized and constructed the local road networks to be matched on the basis of spatial structure characteristic of road networks. Several potential matching local road networks are constituted according to the method of synchronous search of potential matching candidates, and then the corresponding similarity evaluation relationships between road section and road section, local road networks and local road networks are established. The paper puts forward a matching method based on road meshes. Based on the above works, the spatial similarity evaluation model has been constructed by using local road networks as a whole, which considers not only the similarity of individual between road sections, but also the similarity of structure characteristic between local road networks, thus guaranteeing the global logical consistency of the matching results between the selected section to be matched and its adjacent sections. The experimental result has fully proved that the method is effective, with a higher matching rate and accuracy.(5) It is put forward a matching method for multi-scale vector settlement based on global consistency evaluation. Using the road as global constraints and the proximity, similarity and the same orientation which are three factors of Gestalt principle as local constraints, the pattern recognition of settlement cluster has been implemented, thus the distribution characteristics and spatial relationships among settlements are obtained. A local settlement cluster to be matched which is constituted of the selected settlement and its adjacent settlements is regarded as a whole for simultaneous search of potential matching candidate set. As a result, several potential matching local settlement clusters are derived from potential matching candidate sets. The thesis proposed a spatial similarity evaluation model for local settlement cluster, which takes into account not only the similarity of the degree of overlapping, size, orientation and other features of individual settlement, but also the similarity of structure characteristic of local settlement cluster. The experimental result has fully proved that the method can effectively overcome detrimental effects due to positional error, and identify homologous settlements with higher matching accuracy, which has not only matching relationship of 1:1 but also 1:N and M:N simultaneously.更多还原