基于结构感知的多尺度三维网格简化算法

Structure-Aware Multi-Scale 3D Mesh Simplification

摘要:

为解决现有网格简化方法在处理复杂几何和拓扑结构时可能丢失重要信息、以及生成多尺度简化网格效率低等问题，提出了一种基于结构感知的多尺度三维网格简化算法。该方法通过两部分模块组成：基于结构感知的点云采样模块和基于无向距离场的多尺度网格重建模块。基于结构感知的点云采样模块利用学习到的网格结构先验知识，有效地进行点云采样。基于无向距离场的多尺度网格重建模块通过网络预测三维模型的多尺度表面距离场和法线场，以体素级特征表示三维网格，并通过体素表面重建方法实现多尺度三维网格表面的生成。实验结果表明，所提出的方法在网格密闭性、几何形状差异和拓扑结构差异等指标上，都展现出显著的性能优势，能够简化生成更加准确的多尺度网格。

Abstract:

To address the limitations of existing mesh simplification methods， specifically the potential loss of critical information when processing complex geometry and topology， as well as inefficiencies in generating multi-scale simplified meshes， this paper proposes a structure-aware multi-scale 3D mesh simplification algorithm. The proposed method consists of two key components： a structure-aware point cloud sampling module and a multi-scale mesh reconstruction module based on undirected distance fields. The structure-aware point cloud sampling module effectively performs point cloud sampling using learned prior knowledge of mesh structures. Meanwhile， the multi-scale mesh reconstruction module based on undirected distance fields predicts the multi-scale surface distance fields and normal fields of 3D models through neural networks， representing 3D meshes at the voxel level， and enabling the generation of multi-scale 3D mesh surfaces via voxel surface reconstruction methods. Experimental results demonstrate that the proposed approach significantly outperforms existing methods in terms of mesh integrity， geometric shape difference， and topological structure difference， enabling the generation of more accurate multi-scale meshes.