基于主方向高斯映射的旋转面特征提取

为了从点云中准确提取几何特征参数，给出了旋转面特征提取算法.该算法对旋转面主方向映射到高斯球面，生成主方向高斯图像（PDGI）. 对高斯球进行均匀分割，生成一系列小立方栅格，并将高斯球上图像数据点分配到相应的立方栅格中.根据每个栅格中的图像数据点数，对栅格聚类分析，获得一组特征栅格，由这组特征栅格中的图像点来确定大圆所在的平面，从而根据平面法矢确定旋转轴方向.根据旋转面法矢和旋转轴方向，获得旋转轴的定位点，从而确定了旋转轴.并给出了旋转轴的优化方法.计算实例结果表明，该算法能够稳定、准确地提取出旋转面的几何特征参数.

Rotational surface extraction based on principal direction Gaussian image from point cloud

In order to exactly extract the geometric parameters from point cloud, a new algorithm of rotational surface extraction was presented.Principal directions of a rotational surface were mapped to Gaussian sphere in the algorithm,and principal direction Gaussian image(PDGI) was generated.The Gaussian sphere was uniformly partitioned into small 3D grids,and the image points on the Gaussian sphere were assigned to the corresponding 3D grids.A group of 3D feature grids was obtained by cluster analysis technique according to the image points in each grid,then a plane passing through a great circle was fitted to the image points enclosed in the obtained 3D feature grids.The direction of the rotational axis was determined by the normal of the fitted plane.The positional point of the rotational axis was obtained by the direction of the rotational axis and the normal lines of the rotational surface,and then the rotational axis was determined.An optimized method was applied to refine the rotational axis.Experiments show that the algorithm can extract the geometric parameters of the rotational surface accurately and reliably.