激光扫描点云数据的圆柱拟合方法

Cylindrical fitting method of laser scanner point cloud data

针对激光扫描仪点云数据存在较多粗差点导致拟合圆柱不够精确的问题,该文在比较圆柱拟合方法不同特点的基础上,改善了初值任意选取的圆柱拟合方法和基于坐标转换的圆柱拟合方法,对相同的激光扫描仪点云数据进行圆柱拟合,设置相同的收敛条件,每次迭代收敛后,剔除大于3倍中误差的粗差点重新求解,直至没有较大粗差点,最终得到的圆柱参数基本一致。实验表明,改善后的初值任意选取的圆柱拟合方法和基于坐标转换的圆柱拟合方法能够有效剔除激光扫描仪点云数据中的粗差点,且能够满足一般工程的精度要求。

Aiming at the problem that cylindrical fitting is not accurate enough because there are some gross error points among the laser scanner point cloud data,by comparing different characteristics of cylindrical fitting methods,this paper improved the cylindrical fitting method with arbitrary initial values based on coordinate transformation,used the same 3 Dlaser scanning point cloud data and set the same convergence conditions.After each iteration convergence,the points whose correction were more than three times of the mean square error were eliminated,and then restarted calculating.At last,parameters of the cylinder of the two methods were basically the same.The experimental results showed that after improvement the cylindrical fitting method with arbitrary initial values and the cylindrical fitting method based on coordinate transformation could effectively eliminate the gross error points among the laser scanner point cloud data,and could meet the requirements of the general engineering.