一种基于LiDAR点云的建筑物提取方法

发布时间：2018-01-27 15:45

本文关键词： 建筑物提取建筑物轮廓点多种子点区域增长 LiDAR点云　出处：《测绘通报》2017年02期 　论文类型：期刊论文

【摘要】：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。
[Abstract]:Rapid and accurate extraction of buildings from airborne radar point cloud data is a difficult and hot topic in current research. Based on the research and analysis of existing building point cloud extraction methods. In this paper, a building extraction method based on Li DAR point cloud is proposed. Firstly, the initial building contour points are extracted according to the geometric characteristics of the building. Then the local covariance matrix is constructed to calculate the distribution characteristics of the point cloud, and the non-building contour points are eliminated. Finally, DBSCAN clustering algorithm is used to cluster the building contour points. Based on the clustering results, the buffer zone is constructed, and all the building contour points in the buffer zone are taken as the initial seed points. The cylindrical neighborhood is used to grow the multi-seed point area to extract the building point cloud. Through two groups of experiments, the performance of the algorithm is verified by five groups of data. The experimental results show that the method is accurate. It is efficient and applicable to extract multi-story complex building point cloud effectively.
【作者单位】：信息工程大学地理空间信息学院;61618部队;
【基金】：国家自然科学基金(41101396;41001262) 地理信息工程国家重点实验室开放研究基金(SKLGIE2015-M-3-3)
【分类号】：P237;P225
【正文快照】： 建筑物三维模型对于正射影像生成、地形图修测、变化检测、能源和财产管理,以及许多基于位置的服务都至关重要[1],如何生成可靠和准确的建筑物三维模型一直是建筑物三维重建领域研究的热点和难点。激光雷达测量(Li DAR)作为一种通过位置、距离、角度等观测数据直接获取对象表

【参考文献】