Scholarlink

<p>With the rapid development of smart cities worldwide, the demand for 3D Geographic Information Systems (3D GIS) is increasing rapidly. This study focuses on the feasibility of reconstructing Level-of-detail 2 (LOD-2) 3D roof models. Unmanned aerial vehicles (UAVs) imagery was used to generate point cloud, providing three-dimensional coordinate observation information. Using true orthophotos and DSM, the two-dimensional polygons of roof structures were manually digitized to determine their planar boundaries. Least-squares adjustment with blunder error detection were then applied to perform roof plane fitting, ultimately determining the position and distribution of each roof plane in three-dimensional space. In the meantime, the least-squares adjustment with topological constraints was conducted. Corrections for various geometric conditions were applied to further refine the geometric relationships between roof polygons while avoiding topological errors, resulting in an optimized three-dimensional roof model. In the experiment, several buildings with various roof types were evaluated. Accuracy analysis was performed using manually measured 3D coordinates of roof corners. The reconstructed roof models have achieved a planimetric error of 20 centimeters and an elevation error of 15 centimeters, meeting the accuracy requirement of CityGML LOD-2. This confirms the applicability of the proposed roof model reconstruction method.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LOD-2 roof models reconstruction assisted by topological constraints using UAV point cloud

  • 期刊论文

  • 发表日期：

    2026年6月18日

查看全文
  • Meng-Qi Zhou,
  • Iradaf Mandaya,
  • Jiann-Yeou Rau

摘要

With the rapid development of smart cities worldwide, the demand for 3D Geographic Information Systems (3D GIS) is increasing rapidly. This study focuses on the feasibility of reconstructing Level-of-detail 2 (LOD-2) 3D roof models. Unmanned aerial vehicles (UAVs) imagery was used to generate point cloud, providing three-dimensional coordinate observation information. Using true orthophotos and DSM, the two-dimensional polygons of roof structures were manually digitized to determine their planar boundaries. Least-squares adjustment with blunder error detection were then applied to perform roof plane fitting, ultimately determining the position and distribution of each roof plane in three-dimensional space. In the meantime, the least-squares adjustment with topological constraints was conducted. Corrections for various geometric conditions were applied to further refine the geometric relationships between roof polygons while avoiding topological errors, resulting in an optimized three-dimensional roof model. In the experiment, several buildings with various roof types were evaluated. Accuracy analysis was performed using manually measured 3D coordinates of roof corners. The reconstructed roof models have achieved a planimetric error of 20 centimeters and an elevation error of 15 centimeters, meeting the accuracy requirement of CityGML LOD-2. This confirms the applicability of the proposed roof model reconstruction method.