基于3D基元拟合的复杂屋顶点云三维重建_张文元.pdf

引用格式：张文元,陈江媛,谈国新.基于3D基元拟合的复杂屋顶点云三维重建J.地球信息科学学报,2023,25(8):1531-1545.Zhang W Y,Chen J Y,Tan G X.Complex roof structure reconstruction by 3D primitive fitting from point cloudsJ.Journal of Geo-informationScience,2023,25(8):1531-1545.DOI:10.12082/dqxxkx.2023.220927基于3D基元拟合的复杂屋顶点云三维重建张文元，陈江媛*，谈国新华中师范大学 国家文化产业研究中心，武汉 430079Complex Roof Structure Reconstruction by 3D Primitive Fitting from Point CloudsZHANG Wenyuan,CHEN Jiangyuan*,TAN GuoxinNational Research Center of Cultural Industries,Central China Normal University,Wuhan 430079,ChinaAbstract:Geometric and semantic integration of 3D building models are important infrastructure data for smartcity,they are conducive for promoting the refined management and intelligent application of building facilities.However,most of the existing point cloud-based modeling methods focus on the reconstruction of geometricmodels with simple roof structure,and semantic and topological relations are ignored.Moreover,these methodsare sensitive to noise,which are difficult to assure topological consistency and geometric accuracy.To solvethese problems,this paper proposes a 3D primitive fitting algorithm for automatically reconstructing buildingmodels with complex roof structure from point clouds.Firstly,a 3D building primitive library is designed,includ-ing various 3D building primitives with simple and complex roof types.Secondly,an individual building pointcloud input is segmented into multiple planes using RANSAC algorithm.The Roof Topology Graph(RTG)isthen generated according to the relationship of roof planes,and the roof type of point cloud is subsequently rec-ognized by comparison of RTG between point cloud and building primitives.Thirdly,the reconstruction is formu-lated as an optimization problem that minimizes the Point-to-Mesh Distance(PMD)between the point cloud andthe candidate meshed building primitive.The sequential quadratic programming optimization algorithm with nec-essary constraints is adopted to perform holistically primitive fitting,so as to estimate the shape and position pa-rameters of a 3D primitive.Finally,the parameterized model is automatically converted into City GeographyMarkup Language(CityGML)building model based on the prior 3D building primitive.The generated CityGMLLoD2(second level of detail)models are different from mesh models created by conventional building modelingmethods,which are represented with geometric,semantic,and topological information.To evaluate the qualityand performance of the proposed approach,airborne lidar and photogrammetric building point clouds with differ-ent roof structures are collected from public datasets for test.Several building models with complex roof typesare successfully reconstructed by using this approach,and the average PMD of five models is 0.17 m.The pro-posed algorithm is also compared with three other methods.Experimental results indicate that the proposed meth-Vol.25,No.8Aug.,2023第25卷 第8期2023年8月收稿日期：2022-11-28；修回日期：2023-04-15.基金项目：国家自然科学基金项目（41801295）；国家文化和旅游科技创新工程项目（2019-008）。Foundation items:NationalNatural Science Foundation of China,No.41801295;National Culture and Tourism Science and Technology InnovationProject,No.2019-008.作者简介：张文元（1983），男，湖北武汉人，博士，副教授，主要从事三维GIS与文化遗产数字化方面的研究。E-mail:*通讯作者：陈江媛（1997），女，广西来宾人，硕士，主要从事建筑三维建模研究。E-mail:地 球 信 息 科 学 学 报2023年od achieves the best geometric accuracy,because the average PMD of each model is less than that of other meth-ods.Moreover,this automatic primitive fitting method is efficient,and it is also robust to noise and local datamissing.This study demonstrates that the resulting building models can well fit the input point cloud with topo-logic integrity and rich semantic.This method provides great potential for accurate and rapid reconstruction ofgeometric-semantic coherent building models with complex roof condition.Key words:building;complex roof;point cloud;3D reconstruction;3D primitive;CityGML;semantic;model-driven*Corresponding author:CHEN Jiangyuan,E-mail:摘要：几何语义一体化三维建筑物模型是智慧城市建设的重要基础数据，有利于促进建筑设施的精细化管理和智能化应用。当前基于点云的三维重建算法大多关注简单屋顶结构的几何模型构建，忽略了模型的语义表达，且基于数据驱动方法的重建结果容易受噪声影响，存在几何和拓扑错误。为了解决复杂屋顶高精度三维重建难题，本文提出一种基于3D基元拟合的复杂屋顶点云三维自动化重建算法。首先，设计了一套可参数化表达的建筑物3D基元库，包含简单和复杂屋顶。其次，通过点云分割和屋顶拓扑图比较来识别点云对应的基元类型。然后，提出了一种点云与3D基元整体拟合的目标优化函数，采用序列二次规划算法估计基元的正确参数。最后，利用城市地理标记语言（City Geography Markup Language,CityGML）构建几何、语义和拓扑一体化表达的三维模型。采用几种不同屋顶风格的建筑物点云数据进行实验，定性和定量对比分析结果表明本文方法能够高效生成几何和拓扑均正确的CityGML模型，对噪声和局部点云缺失具有一定的鲁棒性，有利于促进几何语义一体化建筑物模型快速自动化构建。关键词：建筑物；复杂屋顶；点云；三维重建；3D基元；城市地理标记语言；语义；模型驱动1 引言建筑物三维建模近年来一直是地理信息系统（Geographic Information System,GIS）、测绘遥感和计算机视觉等领域的研究热点。基于点云的建筑物三维重建不仅可实现自动化建模，而且在建模精度、语义信息提取、几何语义一体化表达等方面都有优势，因而成为了建筑物高精度重建的一种主流技术1-2。近年来国内外大量学者对其进行了研究，杨必胜等3-5对点云数据处理和几何模型重建进展进行了总结。当前基于点云的建筑物三维重建主要分为数据驱动(data-driven)和模型驱动(model-driven)2类建模方法6。基于数据驱动的三维重建是目前建筑物点云三维建模的一种主流方法，主要有区域增长、随机 抽 样 一 致 性(Random Sample Consensus,RANSAC)7、以PointNet8为代表的深度学习网络模型等几类典型算法。赵传等9提出一种区域增长与RANSAC相结合的机载LiDAR点云分割方法，用以分割不同复杂程度的建筑物屋顶面，能有效识别出面积较小的屋顶面。Li等10利用区域增长算法先将点云分割为不规则三角网区域组成的屋顶面片集合，再将其投影到二维格网地图，并引入能量优