点云数据中拉伸面特征的提取

根据拉伸曲面的定义，提出了一种基于点云数据的拉伸面的拉伸方向提取算法，首先，估算点云的主曲率和主方向，在过滤平面数据后计算最小主方向的平均方向，在噪声去除的基础上得到初始拉伸方向；然后，以点云在拉伸方向上的最小投影面积作为目标函数，对初始拉伸方向进行优化得到精确的拉伸方向，最后用几个实例证明了文中方法的正确性、可行性和适用性。

Extruded Surface Extraction Based on Unorganized Point Cloud in Reverse Engineering

With regard to the important part of extruded direction played in feature extraction of an extruded surface in reverse engineering, a method on extracting the extruded direction of extruded surface is presented. Firstly, the principal curvature and direction on each point is estimated based on unorganized point cloud. Then the minimum principal direction is utilized to obtain the initial extruded direction through a sequence of procedures, such as planar data filtering by Gaussian curvature and mean curvature, mean direction calculating about minimum principal direction based on the statistics of directional data, outlier data filtering based on deviation analysis, and so on. Secondly, in order to get a precise extruded direction, a refine algorithm of minimizing the projective area of the point cloud projected onto a plane perpendicular to the extruded direction is presented,and a pattern-searching algorithm to refine the extruded direction is applied. Experimental results illustrate the feasibility of the new algorithm.