基于能量函数基元提取的建筑物LOD2模型自动生成方法  

A method for automatic generating LOD2 building models based on energy function primitive extraction

作　　者：王林[1,2] 陈健龙 刘照亮 刘文轩 WANG Lin;CHEN Jianlong;LIU Zhaoliang;LIU Wenxuan(Hefei Surveying and Mapping Design and Research Institute Co.,Ltd.,Hefei 230061,China;Hefei Engineering Technology Research Center for Geographic Information and Smart City,Hefei 230061,China;Hefei Chaohu Space Surveying and Mapping Technology Co.,Ltd.,Hefei 230061,China;Wuhan Daspatial Technology Co.,Ltd.,Wuhan 430223,China;Engineering Technology Innovation Center for 3D Real Scene Construction and Urban Refinement Governance of the Ministry of Natural Resources,Wuhan 430223,China)

机构地区：[1]合肥市测绘设计研究院有限公司,安徽合肥230061 [2]合肥市地理信息与智慧城市工程技术研究中心,安徽合肥230061 [3]合肥巢湖空间测绘科技有限公司,安徽合肥230061 [4]武汉大势智慧科技有限公司,湖北武汉430223 [5]自然资源部实景三维建设与城市精细化治理工程技术创新中心,湖北武汉430223

出　　处：《测绘通报》2024年第5期103-107,114,共6页Bulletin of Surveying and Mapping

摘　　要：近年来,随着实景三维中国和数字中国的快速发展,建筑物三维模型重建已成为智慧城市建设的重要环节。本文提出了一种全自动建筑物LOD2模型生成框架,旨在解决城市级实景三维场景中建筑物LOD2模型生成的难题。为了自动提取大场景建筑物对象,首先利用场景正射影像,通过图像分割获取建筑物的边界信息;然后使用改进的增强型平面区域分割方法提取高质量的分割平面,并利用混合线性模型对分割结果进行优化,能够更好地处理城市建筑物的局部破损和重建错误等问题。试验结果表明,本文算法生成的建筑物LOD2模型质量和效率更高,能够应对城市建筑模型的复杂情况,得到更稳健的重建结果。In recent years,with the rapid development of 3D real scene(3DRS),the reconstruction of 3D models of buildings has become an important part of smart city construction.This paper proposes a fully automatic framework for generating LOD2 building models,aimed at addressing the challenges of generating LOD2 building models for city-level 3DRS applications.In order to automatically extract building objects in large scenes,this study utilizes orthophoto images and acquires building boundary information through image segmentation.Subsequently,an enhanced plane region growing method is employed to extract high-quality segmented planes,with the results being optimized via a mixed linear model capable of better addressing issues like local damage and reconstruction errors in urban buildings.The experimental results indicate that the proposed method generates higher quality LOD2 building models and adeptly manages complex scenarios in urban building modeling,yielding more robust reconstruction outcomes.

关 键 词：细节层次模型 建筑物重建 区域分割 自动生成 基元提取 

分 类 号：P208[天文地球—地图制图学与地理信息工程]