一种基于点云和拓扑推理的建筑物轮廓线高精度提取方法  

A High-Precision Extraction Method for Building Outlines Based onPoint Cloud and Topological Reasoning

作　　者：陈梦华 张彤蕴 CHEN Meng-hua;ZHANG Tong-yun(Geological and Geographic Information Institute of Hunan Province,Changsha Hunan 410000,China)

机构地区：[1]湖南省地质地理信息所,湖南长沙410000

出　　处：《地矿测绘》2024年第3期7-12,共6页Surveying and Mapping of Geology and Mineral Resources

摘　　要：针对大范围内实景三维模型建筑物轮廓线提取的问题,基于倾斜摄影数据提出了一种轮廓线提取方法。首先,采用区域生长、欧式聚类等方法从倾斜三维模型中得到建筑物的点云数据;然后,采用RANSAC算法及法向约束得到准确的立面;接着,通过基于主基线的拓扑推理生成缺失面;最后,生成完整闭合、平滑吻合的建筑物轮廓线。研究表明,在随机选取的20栋大小不等、样式不一的建筑物上,该方法提取的轮廓线误差低,平面位置中误差为0.05 m,几何形态的长度、宽度以及轮廓面积的均方根误差分别为0.06 m、0.07 m和1.33%。结果显示,该方法简单,提取效率高,可适用于大范围实景三维模型的建筑物轮廓线的提取。A contour extraction method based on oblique photography data was proposed to address the problem of extracting building outlines in large-scale real 3D models.Firstly,the building point cloud data was separated from the oblique 3D model by using regional growth and Euclidean clustering methods.Then,the RANSAC algorithm and normal constraint were used to achieve accurate elevation.Next,the missing surface was generated by topological reasoning.Finally,generate complete,closed,smooth,and consistent building outlines.The research results have showed that the building outlines extracted by this method have quite low errors on 20 randomly selected buildings with different sizes and styles.Specifically,the root mean square error(RMSE)of plane position deviation is 0.05 m,and the RMSE of geometric shape deviation(length and width)and contour areas are 0.06 m,0.07 m,and 1.33%,respectively.Besides,it is simple and efficient,and suitable for extracting building outlines from large-scale real 3D models.

关 键 词：轮廓线提取 倾斜三维模型 RANSAC算法 主基线 拓扑推理 

分 类 号：P231[天文地球—摄影测量与遥感]