摘要
针对建筑物的门、窗、梁、柱等具有规则几何形状的重复模式,提出基于投影点云特征直方图(PPFH)的边界语义分割方法。该方法首先利用PCA方法确定立面的主方向,然后根据点云分布特点,分别进行正立面、侧立面投影得到点云分布直方图,根据直方图的极值点对重复结构边界进行提取和语义分割,最后利用LS-TLS方法进行规则化及误差分析。通过对某建筑物立面窗户提取的算例表明:该方法直接利用重复结构的几何特性,算法简洁,对边缘部分遮蔽和缺失的点云也适用,通过对50个窗户的统计分析,长度和宽度平均误差约6 mm,可广泛用于建筑高精度BIM建模。
For the building’s doors,windows,beams,columns and other repeated patterns with regular geometric shapes,a boundary semantic segmentation method based on Projective Point Feature Histogram(PPFH)is proposed.This method firstly uses PCA method to determine the principal direction of building facade.According to the distribution characteristics between different point clouds,unorganized point clouds are projected onto the front and side elevations to obtain the distribution histogram of point clouds.This paper extracts the repeated structure boundary through the extreme value pair of histogram and complete the semantic segmentation.Finally,the window boundaries are regularized by the LS-TLS method with error analysis.The example captured by Leica P40-for extracting repeated windows of a building facade shows:the method directly uses the geometric feature of repetitive structures,and the process is concise.It can also be used when edges partially obscured or point clouds missing.Through the statistical analysis of 50 Windows’geometric parameters,the average error of length and width is about 6mm,which means it can be widely used for building high-precision BIM modeling.
作者
胡鹏飞
周拥军
HU Pengfei;ZHOU Yongjun(Schoo1 of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiaotong University, Shanghai 200240, China)
出处
《测绘工程》
CSCD
2021年第1期43-50,共8页
Engineering of Surveying and Mapping
关键词
点云
重复模式
语义分割
投影点云特征直方图
point cloud
repeat pattern
semantic segmentation
projection point feature histogram
