车载LiDAR点云中建筑物立面位置边界的自动提取

Automated Extraction of Building Facade Footprints from Mobile LiDAR Point Clouds

提出了一种基于点云特征图像和特征值分析的车载LiDAR点云建筑物立面位置边界的自动提取方法。首先利用车载LiDAR点云数据生成扫描区域的点云特征图像,并通过图像处理手段提取可能的建筑物目标点云;然后对提取的目标点云进行剖面分析和特征值分析,识别建筑物目标;最后对建筑物点云进行平面分割,提取建筑物立面,并对立面点云进行特征值分析,得到建筑物立面与地面交接的三维位置边界。实验结果表明,该方法能快速有效地提取车载LiDAR点云数据中的建筑物目标,同时提取的建筑物立面位置边界与原始点云能准确符合。

We present a novel method for automated extraction of building facade footprints from mobile LiDAR point clouds.The proposed method first generates the georeferenced feature image of a mobile LiDAR point cloud and then uses image segmentation to extract contour areas which contain facade points of buildings,points of trees and points of other objects in the georeferenced feature image.After all the points in each contour area are extracted,a classification based on eigenvalue analysis and profile analysis is adopted to identify building objects from point clouds extracted in contour areas.Then all the points in a building object are segmented into different planes using RANSAC algorithm.For each building,points in facade planes are chosen to calculate the direction,the start point,and the end point of the facade footprint using eigenvalue analysis.Finally,footprints of different facades of building are refined,harmonized,and joined.The experimental results show that the proposed method provides a promising and valid solution for automatically extracting building facade footprints from mobile LiDAR point clouds.