摘要:为了解决在复杂建筑形体的倾斜摄影点云平面分割中，常规的RANSAC平面分割与提取导致的过分割问题，本文结合倾斜摄影群落建筑点云特点，提出一种基于群落内建筑特征统计值的单体建筑关键平面RANSAC优化分割与提取方法。首先，根据群落内建筑平面距离和夹角统计值来合并平面；其次，根据坡度统计值优化竖直平面；再次，根据平面点密度统计值和特征结构进行平面滤除与优化；最后，实现建筑关键平面的优化提取。经验证，该方法对单体建筑的底平面、墙面及屋面等关键特征平面信息的提取准确率为95.80%，召回率为96.21%，能够实现关键特征的精确提取，为后续建筑信息模型自动化构建提供基础数据。

Abstract: Over-segmentation caused by conventional RANSAC plane segmentation and extraction frequently occurs in oblique photographic point cloud plane segmentation of complex building shapes. In this paper, combined with the characteristics of building point clouds in oblique photography, a single building key plane segmentation and extraction method based on the statistical values of architectural features within the community is proposed. Firstly, the planes were merged according to the statistical values of the distance and angle of the building planes in the community. Secondly, the vertical plane was optimized according to the slope statistics. Thirdly, the plane was filtered and optimized according to the statistical value of plane point density and the characteristic structure. Finally, the optimal extraction of key building planes was realized. It has been verified that this method has an accuracy rate of 95.80% and a recall rate of 96.21% for the extraction of key feature plane information e.g. the bottom plane, wall, and roof of a single building, which can provide the basic data for the subsequent automatic construction of building information modeling.