多细节层次的城市三维模型特别是LOD-2细节层次模型是智慧城市构架中的基本要素和模块单元，也是智慧城市的数据基础和信息载体。当前国内外大范围城市三维建模主要以倾斜摄影测量为主，但由于倾斜摄影测量密集匹配点云噪声多、干扰大、自动化程度低、模型精度质量差等问题十分突出，为此，本项目深入研究面向LOD-2的建筑物三维模型自动重建方法，主要研究内容包括：1）空间约束的可靠平面基元提取，保留完整平面结构；2）语义约束的平面基元拓扑关系规则检测与恢复，保证屋顶和立面完整的拓扑重建；3）结构知识引导的二维流形表面自动重建，生成连续无缝的LOD-2模型；4）设计开发原型系统，并选择典型城市场景进行实验分析和验证。本项目旨在为大规模城市精细化三维模型自动重建提供关键的技术支撑，并丰富和发展数字摄影测量理论和方法。

Levels of Detail 3D model of urban buildings, especially LOD-2 model, is the basic element and module unit in the smart city framework, and is also the data foundation and information carrier of the smart city. The current three-dimensional modeling of large-scale city at home and abroad is based on oblique photogrammetry, however, due to the prominent problems of rough noise in dense matching point cloud, low degree of automation and poor quality of model reconstruction by oblique photogrammetry, this project intensive researches the automatic reconstruction method of urban 3D model at LOD-2 level, the main contents include: 1) a reliable planar primitive extraction algorithm by spatial constraints to retain the complete planar structure; 2) the topology detection and recovery of the planar primitive by semantic constraints, to ensure the complete topology reconstruction of the roof and facades; 3) non-manifold surface automatic reconstruction guided by structure knowledge, to generate continuous seamless urban 3D models in LOD-2 level; 4) design and development of prototype system, and conducts the experiments analysis and verification with typical urban scene data. The aim of this project is to provide effective technical support for the large-scale automatic construction of Smart City, and to enrich and develop the theory and method of digital photogrammetry.