提出一种基于直母线族提取与拟合的网格模型直纹面提取方法.首先通过集合误差权排序方法从模型中选择一个可信直母线种子,然后通过局部标架引导搜索邻接直母线,移动标架重复上述搜索过程,直到跨出网格边界或者开始循环搜索.利用“投影”光顺法对齐直母线段族首末端点,再通过定义欧氏6空间下的距离函数,将欧氏3空间下的直线族逼近直纹面问题转换成欧氏6空间下B样条曲线最小二乘拟合问题.为了使逼近的曲面光顺,在曲线拟合过程中引入了能量函数.与其他算法相比,文中方法获得了较强的直母线族的鲁棒性和精确性,并能有效、合理地拟合出光顺直纹面.
We have proposed a novel method to extract ruled surface from mesh model. Based on retrieving and approximating the ruling line sets, our method first selects a trusted ruling line according to sorted error weights. Then neighboring ruling lines are found using the local frame, and the searching of next ruling lines is performed by relocating the local frame until the boundary of an open mesh or the beginning of a cycled searching is reached. A projection method is used to smooth out start and end points of the ruling line sets. With predefined Euclidean 6 space distances, the problem of ruled surface approximation from line sets in Euclidean 3 space can be converted into a least square B-spline curve fitting issue in Euclidean 6 space. For smooth surface approximation, we also add energy function term into the curve fitting process. Compared with other algorithms, examples show that our method can retrieve line sets more robustly and accurately, and can more effectively and rationally fit smooth ruled.