基于法向量估计的激光点云平面特征提取

Plane Extraction from LiDAR Point Cloud Based on Normal Estimation

离散的点云中包含大量的平面特征，其特征在测站拼接、数据压缩、三维建模等方面具有重要应用。本文针对法向量估算方法在过渡边界估算时出现较大误差的现象，首先改进了法向量估算方法，增加了空间距离核函数和法向偏差核函数；然后通过邻域迭代加权估算法向量；最后，通过区域生长方法，选取最小主成分接近零值的点为种子点，并利用邻域内各点法向量和空间距离阈值作为生长条件，提取平面特征。

Discrete point cloud contains a large number of planar features, which play an important role in station stitching, data compression, 3D modeling and so on. In order to reduce the error in transition boundary estimation, the normal estimation method is improved, kernel function of space distance and normal deviation are added, and the normal vector is estimated using neighborhood iteratively reweighted method. Finally, the points whose minimal principal component is approaching zero are se- lected as seed points using region growing method, and the normal vector and space distance threshold of every point in adjacent region are used to extract planar features.