针对目前机载LiDAR点云数据存在的数据组织效率低下以及不利于查询等问题,本文提出了一种基于体元的建筑物提取算法.首先,构建体元模型实现机载LiDAR数据的真三维描述;然后,计算局部邻域曲面拟合残差,将残差最小的体元视作种子体元;最后,根据局部邻域法向量夹角准则来实现种子体元的区域增长,从而获得建筑物点.本文选取ISPRS公开的点云滤波测试数据中的8种复杂场景进行实验,实验结果表明:本文算法不仅原理简单、容易实现,而且具有较好的鲁棒性,不会受地形以及建筑物类型和尺寸的限制,Kappa系数达到80％以上,实现了复杂场景下建筑物的提取.%In view of the present airborne LiDAR point cloud data is the data organization of such problems as low efficiency and not conducive to query,this paper proposes a building extraction algorithm based on voxel.First,the build model implements a true three -dimensional description of airbome LiDAR data;then,it is calculated that the local adjacent surface is fitted with residual difference,and the smallest residues are treated as seed elements.Finally,the regional growth of the seed elements is achieved according to the local neighborhood normal vector angle criterion,and the building points are obtained.This article selects ISPRS public point cloud filter test data of the eight kinds of complex scene experiment,the experimental results show that this algorithm not only principle is simple,easy to implement,and has good robustness,not restricted by terrain and building types and sizes,Kappa coefficient reaches more than 80％,the extraction of complex scenarios for the building.
展开▼
