3D building reconstruction from ALS data using unambiguous decomposition into elementary structures
3D building reconstruction from ALS data using unambiguous decomposition into elementary structures
复制标题
DOI：
10.1016/j.isprsjprs.2016.04.005
复制
发表时间：
2016-08
影响因子：
12.7
通讯作者：
M. Jarzabek-Rychard;A. Borkowski
中科院分区：
文献类型：
--
作者：
M. Jarzabek-Rychard;A. Borkowski
本文的目的是开发一种自动化的方法，使拓扑建筑结构的识别和语义解释。所提出的建模方法的新奇是一个明确的复杂对象分解成预定义的简单的参数化结构，导致在一个拓扑单元的重建没有独立的重叠元素。数据处理链的目标是在LOD 2处生成具有显式拓扑结构和语义信息的完整多面体模型。该算法在机载激光扫描获得的三维点云上进行。所提出的方法结合了基于数据的信息，反映在一个属性的屋顶拓扑图与存储在基本结构库的建筑物的共同知识。为了实现重建精度和可视化方面之间的适当平衡，所实现的库包含一组依赖于结构的软建模规则，而不是严格定义的几何图元。所提出的建模算法首先通过建筑物点云的分割进行屋顶平面提取，然后进行拓扑识别和预定义结构的识别。我们通过分析建模精度和建模细节程度来评估新方法的性能。根据国际摄影测量与遥感学会标准化的验证方法进行的评估表明，算法的完整性在80%以上，正确性在98%以上。
The objective of the paper is to develop an automated method that enables for the recognition and semantic interpretation of topological building structures. The novelty of the proposed modeling approach is an unambiguous decomposition of complex objects into predefined simple parametric structures, resulting in the reconstruction of one topological unit without independent overlapping elements. The aim of a data processing chain is to generate complete polyhedral models at LOD2 with an explicit topological structure and semantic information. The algorithms are performed on 3D point clouds acquired by airborne laser scanning. The presented methodology combines data-based information reflected in an attributed roof topology graph with common knowledge about buildings stored in a library of elementary structures. In order to achieve an appropriate balance between reconstruction precision and visualization aspects, the implemented library contains a set of structure-depended soft modeling rules instead of strictly defined geometric primitives. The proposed modeling algorithm starts with roof plane extraction performed by the segmentation of building point clouds, followed by topology identification and recognition of predefined structures. We evaluate the performance of the novel procedure by the analysis of the modeling accuracy and the degree of modeling detail. The assessment according to the validation methods standardized by the International Society for Photogrammetry and Remote Sensing shows that the completeness of the algorithm is above 80%, whereas the correctness exceeds 98%.