建筑几何正则化约束下的室内点云线段提取

翟若明, 韩贤权, 甘孝清, 邹进贵, 邹双朝, 万鹏, 黎建洲

翟若明, 韩贤权, 甘孝清, 邹进贵, 邹双朝, 万鹏, 黎建洲. 建筑几何正则化约束下的室内点云线段提取[J]. 武汉大学学报 ( 信息科学版). DOI: 10.13203/j.whugis20240384
引用本文: 翟若明, 韩贤权, 甘孝清, 邹进贵, 邹双朝, 万鹏, 黎建洲. 建筑几何正则化约束下的室内点云线段提取[J]. 武汉大学学报 ( 信息科学版). DOI: 10.13203/j.whugis20240384
ZHAI Ruoming, HAN Xianquan, GAN Xiaoqing, ZOU Jingui, ZOU Shuangchao, WAN Peng, LI Jianzhou. Extraction of Line Segments from Indoor Point Clouds under Building Geometric Regularization Constraints[J]. Geomatics and Information Science of Wuhan University. DOI: 10.13203/j.whugis20240384
Citation: ZHAI Ruoming, HAN Xianquan, GAN Xiaoqing, ZOU Jingui, ZOU Shuangchao, WAN Peng, LI Jianzhou. Extraction of Line Segments from Indoor Point Clouds under Building Geometric Regularization Constraints[J]. Geomatics and Information Science of Wuhan University. DOI: 10.13203/j.whugis20240384

建筑几何正则化约束下的室内点云线段提取

基金项目: 

国家自然科学基金(42271447,42001374)

详细信息
    作者简介:

    翟若明,博士,工程师,研究方向为点云数据处理、建筑信息模型生成。ruomingzhai@whu.edu.cn

    通讯作者:

    韩贤权,博士,正高级工程师。hanxq@mail.crsri.cn

Extraction of Line Segments from Indoor Point Clouds under Building Geometric Regularization Constraints

  • 摘要: 针对建筑室内点云三维线段提取方法中存在的线段冗余、完整度低和精度不足等问题,提出了一种融合建筑几何正则化约束的三维线段提取与优化算法。首先,采用点云的区域生长与合并算法完成面片分割,经平面投影提取二维轮廓线,并通过3D-2D反投影得到初始建筑轮廓线段集合。然后,结合建筑几何正则化特性,设计了融合邻接面片相交线、建筑主方向、建筑轮廓共线和建筑结构正交的多重几何约束条件,通过构建的线段优化模型,实现了建筑结构轮廓线的高精度提取,显著提升了其整体的准确性与完整度。该优化过程先分析了面片间的空间邻接关系,对与面片的相交线共线的线段进行方向向量校正,并分析建筑结构主方向与提取轮廓线段的相似度,有效地过滤了冗余的非建筑结构线段,进而通过判断线段间的空间拓扑关系,解决了线段间的不闭合、角度偏移的问题,为既有建筑的室内实景三维重建提供细节丰富的几何结构特征。将该方法在两个室内场景上进行定量分析,结果显示,平均完整度较传统方法提升了6.61%,平均冗余度减少了11.72%,结合对另外两个场景的定性分析,进一步验证该方法在处理复杂建筑结构时的有效性,体现对多样化建筑室内结构的良好适应性和稳健性。
    Abstract: To address the issues of redundancy, low integrity, and insufficient precision in 3D line segment extraction from building indoor point clouds, a 3D line segment extraction and optimization algorithm incorporating building geometric regularization constraints is proposed. Initially, a region growing and merging algorithm is applied to segment point clouds into surface patches upon which 2D profile lines are extracted through plane projection, followed by the 3D-2D back-projection to obtain the initial set of building profile line segments. Next, leveraging architectural geometric regularization features, a multi-constraint model was designed that integrates the building’s principal direction, intersection lines of adjacent facets, collinearity of building contours, and orthogonality of architectural structures. Through this segment optimization model, high-precision extraction of building structure contour lines was achieved, significantly enhancing their overall accuracy and completeness. The optimization process first analyzes the spatial adjacency relationships between patches and corrects the orientation of line segments that are collinear with intersecting lines of patches. It then analyzes the alignment between the building's principal directions and the extracted contour line segments to effectively filter out redundant non-structural segments, while subsequently evaluating the spatial topological relationships between line segments to mitigate the discontinuity and angular deviation, thereby providing detailed geometric features for the indoor 3D reconstruction of existing buildings. Quantitative analysis conducted on two indoor scenes demonstrates a 6.61% improvement in average completeness and an 11.72% reduction in average redundancy compared to traditional methods. Complementary qualitative analysis of two additional indoor scenes further validates its effectiveness in addressing complex building structures, highlighting its robust adaptability and reliability across diverse interior architectural contexts.
计量
  • 文章访问数:  22
  • HTML全文浏览量:  0
  • PDF下载量:  5
  • 被引次数: 0
出版历程
  • 收稿日期:  2024-12-24

目录

    /

    返回文章
    返回