引入法向一致性约束的多尺度邻域漂移点云法向估计算法

Point Cloud Normal Estimation via Multi-Scale Neighborhood Shift with a Normal Consistency Constraint

摘要: 法向量能够描述点云的局部结构，是点云处理中一项重要的几何属性。因此，快速准确地估计法向量是十分关键的。现有基于多尺度邻域漂移的方法通过评估候选邻域的平坦度和空间距离筛选最优邻域，在光滑区域和简单特征上表现良好。然而，该类方法忽略了候选邻域法向与当前点初始法向的一致性，导致在尖锐特征附近容易误选跨越不同曲面的邻域，估计误差较大。因此，本文引入法向一致性约束，对现有算法进行改进，以提高尖锐特征附近法向估计的准确性和可靠性。实验表明，本文方法能够提升最优邻域的筛选效果，有效降低尖锐特征附近的法向估计误差，对形状较为复杂的模型表现出更好的鲁棒性。

Abstract: Normal vectors describe the local structure of point clouds and are an important geometric attribute in point cloud processing. Therefore, a fast and accurate normal estimation is crucial. Existing methods based on multi-scale neighborhood shift select the optimal neighborhood by evaluating the flatness and spatial distance of candidate neighborhoods. These methods perform well on smooth regions and simple features. However, such methods ignore the consistency between the normals of candidate neighborhoods and the initial normal of the current point. This easily leads to the incorrect selection of neighborhoods that cross different surfaces near sharp features, resulting in large estimation errors. Therefore, this paper introduces a normal consistency constraint to improve existing algorithms, thereby enhancing the accuracy and reliability of normal estimation near sharp features. Experimental results show that the proposed method can improve the effectiveness of the optimal neighborhood selection and effectively reduce normal estimation errors near sharp features. Furthermore, it exhibits better robustness to models with more complex shapes.