基于机载LiDAR数据的建筑物轮廓提取

建筑物轮廓作为建筑物三维重建的重要元素，在建立智慧城市和数字城市中至关重要。本文针对从机载激光雷达点云中提取建筑物轮廓数据处理的点云滤波、建筑物屋顶面提取、建筑物轮廓提取，以及提取精度评定各环节存在的一些问题，提出了一种综合区域生长改进算法、三维Hough变换算法和α-shape算法的建筑物轮廓提取方法。该方法在对机载LiDAR点云数据去噪的基础上，首先利用改进的区域生长算法滤波地面点，并基于地物点到地面的归一化高程特征通过高度阈值去除高度较为低矮的地物点；再基于三维Hough变换算法从剩余建筑物和高大树木点云中提取建筑物平面；最后使用α-shape算法提取建筑物的轮廓信息。对使用RIEGLVQ-1560i机载激光雷达测量系统扫描的某城区点云数据进行计算，通过匹配度、形状相似度和位置精度等评价指标对提取的建筑物轮廓进行精度评定。结果表明，综合区域生长改进算法、三维Hough变换算法和α-shape算法的建筑物轮廓提取方法可以准确提取建筑物的轮廓信息，对于大范围的建筑物轮廓提取具有稳定性和普遍适用性。

Building outline extraction based on airborne LiDAR data

As an important element of 3D reconstruction of buildings, building contours are crucial in building smart cities and digital cities. Aiming at some problems in the extraction of point cloud filtering, building roof surface extraction, building contour extraction and extraction accuracy evaluation from the airborne LiDAR point cloud, this paper proposes a comprehensive region growth improvement algorithm, three-dimensional Hough transform algorithm and α-shape algorithm for building contour extraction method. Based on the denoising of airborne LiDAR point cloud data, the method uses an improved region growing algorithm to filter the ground points and removes the lower height features through the height threshold based on the normalized elevation features of the ground point to the ground. Then, based on the three-dimensional Hough transform algorithm, the building plane is extracted from the remaining buildings and the tall tree point cloud, and finally the contour information of the building is extracted using the α-shape algorithm. The point cloud data of a certain urban area scanned by the RIEGLVQ-1560i airborne LiDAR measurement system is calculated, and the accuracy of the extracted building outline is evaluated by the evaluation indexes such as matching degree, shape similarity and positional accuracy. The results show that the integrated region growth improvement algorithm, three-dimensional Hough transform algorithm and α-shape algorithm for building contour extraction method can accurately extract the contour information of buildings, and have the characteristics of stability and universal applicability for wide-area building contour extraction.