基于RANSAC和TLS的点云平面拟合

Plane fitting of point clouds based on RANSAC and TLS

针对点云平面拟合过程中出现的异常值及误差的问题,提出一种将随机采样一致(random sample consensus,RANSAC)算法与整体最小二乘法(total least squares,TLS)相结合的点云平面拟合方法。利用随机采样一致算法剔除异常值,利用整体最小二乘法对剩余有效点进行平面拟合,计算模型参数。实验结果表明,该方法与传统的特征值法、最小二乘法相比,能提高参数的估算精度,更适合对含有不同异常值及误差的点云数据进行拟合,是一种稳健的平面拟合方法。

To resolve the problems of outliers and errors in the process of point cloud plane fitting, a method combining the random sample consensus algorithm and total least squares was proposed. Random sample consensus algorithm was used to eliminate outliers, and the total least squares were used to do plane fitting for the remaining valid points. Experiments results verify that the method is more adaptive to fit plane in various point clouds compared to the methods such as least squares and eigenvalue method, and it can improve the estimate precision of the model parameters and it can steadily get fine planar parameters with good robustness.