语义辅助的多维空间数据智能集成方法

Heterogeneous spatial data integration is the key technology to support the interdisciplinary collaborative research of geography. Among them, multidimensional spatial data integration is the main method of conflating global/regional scale spatial data with local finer 3D model data. At present, the relevant theories and methods are not fully consider the role of artificial terrain/building semantical rules in multidimensional spatial data integration. In this proposal, we are going to classify the rules of the above spatial entities' implicit semantics, excavate the spatial/non-spatial metaphor available, and construct the semantic rules base. We will investigate the semantic rules, which exist in and implied by conflating the terrain and the buildings, between buildings, and exterior/interior of same building. The multidimensional spatial data integration modeling algorithm, the non-spatial metaphorical constraint objective function and the constraint condition construction method and the optimization algorithm are studied based on the above consistency rule. In the above research process, the effect of scale effect in multidimensional spatial data integration is studied by integrating experiments with multi-scale/multi-resolution/multi levels of details spatial data. The successful implementation of this proposal will help to develop semantic-aided multi-dimensional spatial data integration theory and method, and provide data environment support for cross-domain applications such as integrated navigation, emergency rescue, disaster prevention and mitigation, virtual geographic environment and so on.

异构空间数据集成是支撑地理学跨学科综合协同研究的关键技术，其中，多维空间数据集成是全球/区域尺度建模与局部三维精细化应用结合的主要方法。目前相关理论和方法的发展未能充分考虑人工地形/建筑物语义隐含规则在多维空间数据集成建模中的作用。本课题拟对上述实体语义隐含的规则进行机器学习与分类，挖掘其中可用的空间/非空间隐喻，构建语义规则库；研究语义规则隐含的地形与建筑物、建筑物之间及其内外部语义、拓扑、几何一致性关系；研究基于上述一致性规则的多维空间数据集成建模算法，非空间隐喻约束目标函数与约束条件构建方法以及优化求解算法。在上述研究过程中，通过多尺度/多分辨率/多细节层次空间数据集成对比实验，研究尺度效应在多维空间数据集成中的影响。本课题的成功实施有助于发展语义辅助的多维空间数据集成理论与方法，对室内外一体化导航、应急救援、防灾减灾、虚拟地理环境等跨领域应用提供数据环境支撑。

本研究结合目前地理空间信息领域多源数据的现状，特别是地形、道路、建筑物、植被等地形地物语义对象的矢量、栅格、点云等多源数据共存的情况，针对以矢量与栅格所表达的地形与道路数据、以CityGML模型和BIM模型及点云等表达的建筑物数据、以矢量与激光点云数据等所表达的多类型场景数据等，在地形地物语义隐含的规则库构建、规则约束下的数据转换、地形与道路数据集成与融合、基于点云的地物语义分类与分割、基于点云的目标特征提取与建模等方面进行了研究，设计了建筑物语义约束下的几何一致性规则，提出了语义辅助的建筑物真三维空间数据模型，设计了语义辅助的矢量道路与地形增强建模方法，提出了行道树等典型植被的激光点云分类/分割方法，以及建筑物特征提取与建模方法等。上述研究在一定程度上发展了传统矢量/栅格数据的集成、二维/2.5维/三维数据的融合、激光雷达点云数据智能处理等领域的理论，相关模型与技术方法可推动上述多源数据在实景中国、智慧城市、应急反恐等领域的应用。