建筑物立面点云直线段特征提取方法

Straight-Line-Segment Feature-Extraction Method for Building-Facade Point-Cloud Data

针对现有从建筑物立面点云数据中检测提取直线段特征的方法存在漏检现象严重和准确程度不高的问题,提出一种基于切片的建筑物立面点云直线段特征提取方法。首先对建筑物点云姿态进行调整,使其走向与Y坐标轴一致,然后沿三个坐标轴方向对点云进行切片并在切片上提取特征点;之后分别对三个方向提取的特征点基于圆柱体生长的方式进行直线段聚类;最后采用残差1范数最小进行聚类特征点的直线段拟合及对直线段端点进行调整和优化。采用多组实验数据对本方法进行验证,实验结果表明:本文方法的直线段提取精度为点云中平均点间距的1/2;与基于平面分割和图像检测的方法相比,本文方法提取直线段的精确率平均提高了2.4%,召回率平均提高了48.1%,可以更加准确有效地从建筑物立面点云数据中提取直线段特征。

In this study, we propose a straight-line-segment feature-extraction method for the building-fa?ade point-cloud data based on slicing to improve the existing method of detecting and extracting the straight-line-segment features from the building-fa?ade point-cloud data, which exhibits problems of missed detection and less-than-optimal accuracy. Further, the point cloud is sliced along the three coordinate axes after adjusting the point-cloud attitude of the building to ensure that its orientation is consistent with the Y-coordinate axis. Then, the feature points on each slice are extracted, and straight-line-segment clustering is applied to the extracted feature points based on the cylinder growth method. Finally, the straight-line-segment fitting of the feature points is performed using the 1-norm minimum residual algorithm, and the endpoints of the straight line segment are adjusted and refined. Subsequently, we validate the proposed method by applying it to several sets of experimental data;the experimental results exhibit improved accuracy, precision, and recall. The extraction accuracy of the straight line segment is half the average point spacing in the point cloud. The precision of the proposed method for extracting straight line segments is increased by 2.4% on average than that of the plane segmentation and image detection methods, whereas the recall is increased by 48.1% on average. Thus, our proposed method can accurately and effectively extract straight line segments from the building-facade point-cloud data.