基于高斯映射的柱面与锥面点云拟合

由于点云数据中存在噪声的影响，采用基于点云数据的拟合算法对圆柱面和圆锥面拟合时进行，获得的拟合初值不稳定，从而严重影响了最终拟合曲面的准确性。针对这一重要问题，提出了一种高斯映射方法来获得良好的拟合初值算法。该算法将柱面和锥面的单位法矢映射到单位球上，形成高斯映像。利用聚类分析方法剔除高斯映像数据中的噪声，获得干净的高斯映像数据，利用这些干净的高斯映像数据获得准确的轴线方向和轴线位置，确定了较好的拟合初值，最后采用非线性最小二乘方法获得准确的柱面和锥面。

Fitting Point Cloud to Cone and Cylinder Based on Gaussian Image

As for the point data are affected by noise in point cloud, the initial estimates are usually unstable that often leads to failure or incorrect result in the process of fitting point cloud to cone or cylinder. As regard to this important problem, a Gaussian image based algorithm is proposed to obtain good initial estimates. In this method, the Gaussian image of a cone or a cylinder is first obtained and then cluster analysis is applied to removing the noise data and acquiring the clean data. The accurate direction and position of the rotational axis, as one of the most important parameters of initial estimates, can be calculated from these clean data. At last, a nonlinear least-squares method is used to refine the solution.