本文以表面连续的折线型边缘和小圆弧屋脊型边缘为研究对象, 提出了一种基于扫描点云数据的线轮廓提取方法。该方法通过构造万向切片从点云数据中提取包含线轮廓点的截面数据, 根据截面数据上线轮廓点两侧的分布形式, 定义了基于法向夹角分布规律的模式向量, 将截面数据的相邻点法向夹角序列进行等元素划分, 计算各组与模式向量的欧氏距离, 在欧氏距离最小的组内提取线轮廓点。为验证方法的准确度, 分别对直线型折线轮廓、弯曲折线轮廓和4个圆形屋脊轮廓进行试验, 并将提取的线轮廓点进行最小二乘拟合, 以线轮廓点相对拟合曲线的偏差评价方法的准确度。提取的直线轮廓点和曲线轮廓点的拟合标准偏差分别为0.076 mm和0.047 mm; 4个圆形轮廓点的拟合标准偏差不大于0.1 mm, 圆半径相对三坐标测量结果的偏差不大于0.1 mm。模式向量法适用于提取折线型边缘和小圆弧屋脊型边缘上的轮廓点数据, 具有计算简单, 适用性强的特点。当扫描仪准确度优于0.03 mm时, 模式向量法的准确度在0.1 mm的量级。

In this article, an extraction method of line contour points from point cloud data (for a fold-line edge and a ridge edge with a small chamfer) is proposed. In this method, multiple-direction slices were constructed to obtain section data containing line contour points. Then, according to the form characteristic of the section data, a pattern vector was constructed by using the normal angle of two adjacent points to represent the characteristic. The section data was divided into many small groups, each with the same number of components. Euclidean distances between the pattern vector and each small group were calculated, and the group with the smallest Euclidean distance was located for picking out the line contour point. To determine the accuracy of the proposed method, examples of three kinds of edges were used: a straight-line edge, a curve edge, and four circles. The extracted line contour points were fitted to a straight line, a three-order polynomial, and circles, by the least square fitting method. The deviations of the line contour points to the fitted lines were calculated. The standard deviations of the straight-line and curve edges were 0.076 and 0.047 mm, respectively. The standard deviations of the four circles were smaller than 0.1 mm. The radius deviation of the four circles to the reference data, measured by a coordinate measuring machine, was within 0.1 mm. This pattern vector method is applicable for extracting line contour points on a fold-line edge and a ridge edge with a small chamfer. When the scanning accuracy is higher than 0.03 mm, the pattern vector method ensures an accuracy of 0.1 mm.