中文摘要：

针对常用的平面拟合方法在点云数据存在粗差或异常值扰动时，存在拟合结果不稳定的缺点，提出了一种稳健的点云数据平面拟合方法．该法以特征值法为基础，通过利用一定的准则删除点云数据中的粗差或异常值，从而获得稳健的平面参数估计值．在实验中，分别利用最小二乘法、特征值法和该稳健特征值法对点云数据进行拟合，结果显示该法能克服异常值的影响，得到可靠的平面参数估值，具有稳健性．

英文摘要：

The results of planar parameters estimation are not accurate by traditional plane fitting methods to point clouds,because the gross error and outliers are not considered. This paper presents a robust method for fitting local plane to point clouds. The proposed method, named robust eigenvalue method, is based on eigenvalue method. Through deleting outliers from point clouds,a robust solution to plane fitting parameter can be obtained.Analytical simulation experiment was oonducted, and a comparative study was also made between the proposed and traditional methods such as least square method and eigenvalue method. The results show that the method can overcome the influence from outliers,and improve the reliability of parameter estimation.