基于点对距离差异性的点云法向估计算法

Point Cloud Normal Estimation Algorithm Based on Point Pair Distance Differences

摘要: 法向量作为点云数据的关键属性，在众多算法中发挥着重要作用。然而，由于点云数据易受噪声干扰、存在离群点以及非均匀采样等问题，如何精准地估计尖锐特征点的法向量仍是该领域极具挑战性的难题。本文提出一种基于点对距离差异性的点云法向估计算法，核心是通过所有相邻点对对局部切平面进行投票的一种方案。本算法深入分析点云中点对之间的距离特征，利用高斯权函数量化距离信息，在考虑点对法向一致性基础上加入点对之间距离信息参与投票过程，从而构建出有效的点云法向估计模型。实验结果表明，相较于传统点云估计算法，该算法的法向估计误差得到减少，并且能准确地恢复出复杂模型的细小特征，有效地克服噪声和非均匀采样。

Abstract: As a crucial attribute of point cloud data, normal vectors play a significant role in numerous algorithms. However, accurately estimating normals at sharp feature points remains a challenging problem in this field due to inherent issues in point cloud data such as noise susceptibility, presence of outliers, and non-uniform sampling. This paper proposes a novel point cloud normal estimation algorithm based on distance differences between point pairs, whose core innovation lies in a voting scheme where all neighboring point pairs collectively determine the local tangent plane. The algorithm thoroughly investigates distance characteristics between point pairs in cloud data, employs Gaussian weighting functions to quantify distance information, and constructs an effective normal estimation model by incorporating both distance information and normal consistency into the voting process. Experimental results demonstrate that compared with traditional point cloud estimation algorithms, the proposed method achieves reduced normal estimation errors while accurately recovering fine features of complex models, effectively overcoming challenges posed by noise and non-uniform sampling.