中文摘要：

针对传统平面拟合算法难以拟合包含异常值点云的问题，提出结合特征值法的随机抽样一致性（RANSAC）平面拟合算法。随机选取三个点云数据直接计算平面，选择阈值统计在此平面上的内点数量，多次重复求得包含最多内点的平面，并以这些内点以特征值法进行平面拟合得到所求平面方程。对各种包含误差及异常值的平面点云进行拟合计算，并与传统算法进行比较，将其应用于双目重构得到的隧道开挖掌子面岩体三维数字模型中节理面点云平面拟合。实验结果表明该方法可以很好地适应各种误差和异常值的情况，稳定地得到较好的平面参数估计值，是一种鲁棒的平面拟合算法。

英文摘要：

It is hard to fit points cloud with exceptional points by classical plane fitting algorithm.A robust plane fitting algorithm based on RANSAC and eigenvalue method is advanced to solve this problem.Three points are selected randomly to compute plane parameters,and the number of inner points is counted with proper threshold.This processing is repeated for certain times and points group with the most inner points is selected,with these points eigenvalue method is used to plane fitting.Fitting experiments are carried out with various kinds of error and exceptional points,after that it is used in rock mass discontinuity points cloud on 3-D rock tunnel face digital modal,which is generated with binocular system.Conclusions are drawn that compared with classical methods the proposed method can adapt various situation of error and exception and get fine planar parameters steadily.It is a robust plane fitting algorithm.