基于平面特征的地面雷达点云配准算法

Point cloud registration algorithm for terrestrial LiDAR based on planar features

针对地面LiDAR点云配准中不同坐标系点云数据存在对应的平面特征不同的问题,文章提出了一种基于总体最小二乘的地面LiDAR点云数据配准算法:通过对分割后的点云数据平面拟合,得到相应法向量;根据不同坐标系中LiDAR点云数据对应的平面法向量,利用反对称矩阵和罗德里格矩阵的性质,用3个独立参数代替3个旋转参数,采用总体最小二乘法建立旋转矩阵解算模型;采用总体最小二乘法确定平移参数的计算公式;最后根据转换后特征点云与对应平面点云的重复情况,给出了配准模型的精度公式。实验结果表明该方法精度较高,可以取得较好的点云配准效果,适合于含有大量重复平面特征的点云数据的配准。

Aiming at the problem that in the terrestrial LiDAR point cloud registration,the corresponding planar characteristics of the data in different coordinate systems are different,the paper proposed a registration method based on Total Least Squares（TLS）：the normal vectors of the planes were obtained after fitted from the segmented point cloud data;by taking advantage of the property of the anti-symmetric matrix and Rodriguez matrix,three angle parameters were replaced by three independent elements using the corresponding plane normal vectors of LiDAR point cloud data in different coordinate systems;then the rotation matrix calculation model and the translation parameters formula were established using TLS;finally,the error equation of rotation transformation was deduced by the duplication between the transformed planar points and the corresponding planes,and the registration accuracy was analyzed and compared with ICP.Experimental result showed that the method could be applied in the point cloud registration containing a lot of repetition planar characteristic with its high matching precision.