提出一种对逆向工程网格噪声鲁棒的分水岭分割算法.该算法在计算网格离散曲率时,针对曲率计算对网格噪声特别敏感的问题,根据拟合曲面的曲面误差估计,动态地调整拟合曲面的顶点个数,提高了曲率计算的精确性,增强了基于曲率的分水岭算法对噪声的鲁棒性;通过后续的标识、聚类和分割后处理方法,提高了算法的分割精度和效果.该算法在大量的噪声网格模型上获得了较好的分割结果,适用于逆向工程中的二次曲面识别和NURBS曲面逼近。
This paper proposes a watershed mesh segmentation algorithm which is designed to be robust to mesh noise. The calculation of discrete curvature will evaluate the error of approximating surface and adjust the number of approximating vertices dynamically. A series of subsequent marking and clustering heighten the effect and precision of mesh segmentation in existence of noise. The algorithm achieves satisfying results on a considerable of models corrupted by noise. The segmentation results are proper for quadric surface identification and NURBS approximating in reverse engineering.