基于Cook距离的三维激光扫描点云平面拟合

PLANE FITTING OF POINT CLOUD FROM 3D LASER SCANNING BASED ON COOK DISTANCE

针对存在较多误差和粗差的点云数据,直接用最小二乘拟合局部平面的精度和可靠性不高的问题,提出了基于Cook距离的最小二乘拟合方法,先从大量的点云数据中优化选取强影响点,再通过最小二乘方法获得准确的拟合平面。试验表明,该方法拟合出的平面与真实平面较接近,具有较高的精度和稳定性,可以大幅度缩减点云并建立更精确的模型。

Traditional least-squares can not ensure high accuracy and stability for plane fitting when many out- liers and error are mixed in original data. To solve this problem, a least-squares method based on Cook distance is proposed. The method determines strong impact points by Cook distance value from point cloud data optimally and then accurately fitted plane can be achieved by least-squares. The result of experiment and analysis shows that the fitted plane is closer to the real plane. The method can improve accuracy and stability of the process of point cloud reducing and model building effectively.