高级检索
王鹏飞, 徐敏峰, 辛士庆, 严冬明, 屠长河. 鲁棒的水密流形网格修复[J]. 计算机辅助设计与图形学学报, 2024, 36(7): 1047-1056. DOI: 10.3724/SP.J.1089.2024.19892
引用本文: 王鹏飞, 徐敏峰, 辛士庆, 严冬明, 屠长河. 鲁棒的水密流形网格修复[J]. 计算机辅助设计与图形学学报, 2024, 36(7): 1047-1056. DOI: 10.3724/SP.J.1089.2024.19892
Wang Pengfei, Xu Minfeng, Xin Shiqing, Yan Dongming, Tu Changhe. Robustly Watertight Manifold Surface Repair[J]. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(7): 1047-1056. DOI: 10.3724/SP.J.1089.2024.19892
Citation: Wang Pengfei, Xu Minfeng, Xin Shiqing, Yan Dongming, Tu Changhe. Robustly Watertight Manifold Surface Repair[J]. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(7): 1047-1056. DOI: 10.3724/SP.J.1089.2024.19892

鲁棒的水密流形网格修复

Robustly Watertight Manifold Surface Repair

  • 摘要: 针对未经修复的网格模型一般存在非流形结构,常带有孔洞、法向不一致、自交等缺陷,很难直接应用到后续基于网格的应用中的问题,提出一种保持输入网格特征的鲁棒水密流形网格修复算法.首先利用Manifoldplus算法和卷绕数(winding number)构建能够区分输入网格内外且逼近输入网格的水密流形引导曲面;然后利用引导曲面计算受限Voronoi图(restricted Voronoi diagram,RVD);再通过对偶得到受限三角剖分(restricted Delaunay triangulation,RDT);将非流形问题分解到RVD和RDT计算过程中,保证计算的RDT即为修复后的水密流形网格;最后在原始网格边中添加辅助点,保持原始网格特征.基于Windows 10平台,在ModelNet10公开数据集上进行实验的结果表明,所提算法在输出网格的平均精度为1.54×10-6,与Manifoldplus算法相当;但是当输入的模型包含孔洞时,Manifoldplus算法无法将孔洞合理地填补,而该算法能够合理地填补孔洞.

     

    Abstract: Raw mesh is often non watertight manifold, and contains a variety of defects, such as holes, inconsistent surface normal, self-intersection and so on, which is difficult to be directly applied to subsequent applications. Therefore, a robust watertight manifold mesh repair method is proposed. Firstly, the Manifoldplus and winding number are used to build a watertight manifold guiding surface that approximate the input mesh. Then, the restricted Voronoi diagram (RVD) is computed on guiding surface and restricted Delaunay triangulation (RDT) is obtained through duality. By decomposing the non manifold problem into the RVD and RDT, it is ensured that the computed RDT is watertight manifold mesh. Finally, auxiliary points are added to the original mesh edges to maintain the feature of original mesh. Based on the Windows 10 platform, the comparison experiment on the ModelNet10 public dataset shows that the average accuracy of the proposed algorithm is 1.54×10-6, equivalent to the Manifoldplus method. However, when the input model contains holes, manifestplus can not fill the holes reasonably. In contrast, the proposed algorithm can fill the holes reasonably.

     

/

返回文章
返回