一种改进的最大一致性点云几何基元拟合算法

An Improved Maximum Consistency Geometric Primitives Fitting Algorithm for Point Cloud

摘要:

基于MCMD_Z(maximum consistency with minimum distance and robust Z score)算法思想，提出了一种稳健的且适用于平面、二次曲面(球、圆柱、圆锥)基元高精度拟合算法.算法依据距离和最小准则，从含有粗差的点集中选取最佳点子集拟合可靠模型初值，并采用稳健Z分数方法循环剔除粗差；对剔除粗差后的保留点集采用加权最小二乘迭代方法拟合.实验表明，对粗差含量较高的点云数据，该算法均能有效剔除粗差、拟合出高精度的几何基元.

Abstract:

Based on the idea of MCMD_Z algorithm, this paper presented a robust high precision fitting algorithm for plane, quadric surface primitives(sphere, cylinder, cone). According to the minimum sum of distance criteria, the algorithm obtained the best subset from the point cloud to fit the reliable initial value of the geometric primitive, removed the outliers cyclically using the robust Z score method, and fitted the inliers by using the weighted least square iteration method. Experimental results show that this algorithm can effectively remove outliers and precisely fit the geometric primitive in the point cloud with high content of outliers.