In common plane fitting methods for point clouds，the results of planar parameter estimation are not always accurate when the gross errors and outliers are included. To overcome this shortcoming, ＲANSAC ( random sample consensus) algorithm is proposed combined with least square method． The ＲANSAC algorithm is adopted to detect and eliminate the outlier points，and least square method makes plane fitting for the remaining inner valid points. Analytical simulation experiments have been conducted. Comparative results between our method and traditional methods，such as least squares and eigenvalue method，are provided. The proposed method is also used for solving the point clouds problems with varying scale of errors and outliers. Calculation results verify that the method is adaptive to plane fitting in various point clouds，and it can steadily get fine planar parameters with good robustness.