一种基于LiDAR点云的建筑物提取方法

从机载雷达点云数据中快速准确提取建筑物是当前研究的难点和热点。在对现有建筑物点云提取方法充分研究和分析的基础上,本文提出了一种基于Li DAR点云的建筑物提取方法。首先根据建筑物的几何特性提取初始建筑物轮廓点;然后构建局部协方差矩阵计算点云分布特征,剔除非建筑物轮廓点;最后利用DBSCAN聚类算法对建筑物轮廓点聚类,以聚类结果为基础构建缓冲区,以缓冲区内所有建筑物轮廓点为初始种子点,采用圆柱体邻域进行多种子点区域增长,实现建筑物点云的提取。通过两组试验,共5组数据验证本文算法的性能。试验结果表明,该方法能够准确、有效地提取多层复杂的建筑物点云,效率高,且具有一定的适用性。

A Method of Extracting Building Based on LiDAR Point Clouds

Extracting buildings from LiDAR cloud points quickly and precisely is a difficult and hot spot in current research. On the basis of analysing existing building extraction methods sufficiently, a building extraction method is proposed based on LiDAR point cloud. Firstly, initial building contour points are extracted according to geometric characteristics of buildings. To eliminate false build-ing contour points, distribution of LiDAR point cloud is then calculated by constructing the local covariance matrix. Finally, buffer zones are built based on clustering result of building contour points by using DBSCAN clustering algorithm, all building contour points in each buffer zone are selected as initial seed points, and multi-seed points region growing process, adopting a cylinder neighborhood system, is applied to achieve extracting buildings quickly and accurately. Two sets of experiments, concluding a total of five datasets, have been realized to verify all aspects performance of the algorithm. Experimental results show that the method can extract multistorey and complex buildings accurately and effectively, and performs high efficiency and strong applicability.