支持自动综合的多尺度三维建筑组织与管理

多尺度数据的有效组织与管理是实现三维建筑自动综合及多尺度应用的基础。本文对空间数据综合与多尺度表达数据管理方法的研究现状进行了分析。根据三维建筑自动综合的特点,本文提出了一种基于R树的多尺度三维建筑空间索引模型——GAMR树（Generalization Area Multi-representation R Tree）;对GAMR树的定义和结构进行了详细描述,设计了顾及GAMR树高度与所选长度阈值关联关系的索引构建方法。针对三维建筑特点,设计了一种支持多尺度区域对象完整查询的检索方法,并根据建筑特征的可辨性,实现了顾及视线角度对投影长度影响的三维建筑模型多尺度可视化方法。实验证明,GAMR树能很好地适用于三维建筑多尺度模型的组织与管理,对三维建筑几何模型的自动综合具有重要意义。

Organization and Management of Multi-scale 3D Buildings for Generalization

With the rapid development of geographical information acquisitions methods, 3D city models are widely used now. Level of detail (LOD) is used to improve the performance of three-dimensional (3D) visualization. Management of multi-scale building data is important in 3D building generalization, and many scholars have studied the indexing model of multi-scale spatial data for 2D map. In order to satisfy the demands of massive spatial data in 3D city modeling, a new spatial indexing model named as GAMR tree (Generalization Area Multi-representation R Tree) is proposed to deal with the multi-scale model of 3D building. Based on the characteristics of 3D building generalization, impact factors for constructing the 3D building generalization areas are analyzed in this paper. Next, the generalization areas are divided into four types which are denoted as big city, city, town and village. As a result, different generalization methods are adopted for the generalization of different areas. In this research, GAMR tree is defined based on the structure of R-Tree while assigned with different node types. Two indexing trees, which are named respectively as the generalization R tree and the child tree of generalization area, are included in the new indexing model. The leaf node of the child tree is consisted of two items, one is the MBR that representing the building ground area, and the other is the corresponding 3D building model for that area. While the branch node is consist of a MBR that accounts for the total area of its child nodes, and pointers that connect to its child nodes. The children trees are generated by a bottom-up method, and the height of a tree is determined by the threshold calculated in generalization. Multi-scale querying method of GAMR tree is furthermore proposed in this paper, and the steps of a rectangle querying instance are presented. Considering the variations of projected length caused by different sight directions, the corresponding LODs of building models that organized by GAMR tree is well determined and chosen to be displayed on the screen, so that the projection of building features can be recognized by users. Generally, the experiment has proved that this spatial indexing tree works well in 3D building generalization and multi-scale representation.