基于无人机载激光雷达数据的建筑物自动提取算法

摘要

针对机载LiDAR散乱点云数据量大、目标分类困难的问题，提出一种实现LiDAR点云数据中地面点云、植被点云和建筑物点云的全自动分类算法，并根据提取的建筑物点云数据自动提取建筑物轮廓和中心点坐标。具体理论算法为：首先基于渐进三角网的点云数据滤波算法分离出地面点云数据，然后根据植被点云法向量的各向异性采用模糊C均值聚类（FCM）方法分离植被点云和建筑物点云。对于分类后的建筑物点云，利用拓扑聚类的方法，对每个建筑物进行识别并提取轮廓和中心点。将该算法应用于栾川协心小流域的山洪灾害调查评价居民户建筑物位置和高程自动提取。应用案例表明：该方法提取速度快，提取精度较高，且适用于山丘区机载雷达数据建筑物提取和植被分类。%Airborne-lidar faces chal enges to process large amount of scattered point cloud data and classification. This paper introduces an automatic classification algorithm of LiDAR point cloud data processing, which could automatical y classify the ground data, vegetation data, as wel as extract data for the contour of buildings and center point coordinates according to extracted building point cloud data. The algorithm based on the incremental triangulation of the point cloud data filtering algorithm to separate the data of ground points, and considering the anisotropy of the normal vector of vegetation data, taking FCM fuzzy clustering method to separate the data of vegetation and buildings. After classification building point cloud, the outline and center point for each building were identified using topological clustering method. The algorithm was applied to flash floods disaster in smal watershed in Luanchuan County to extract locations and heights of residential buildings. The results show that the method can process data rapidly with high precision; therefore it is suitable for extracting date from building and surface vegetation classification in hil y areas.