基于三维点云模型的特征线提取算法

针对以往算法存在无法区分尖锐和非尖锐特征点、提取的特征点与视角有关、特征点未连线等问题, 提出一种基于高斯映射和曲率值分析的三维点云模型尖锐特征线提取算法。该算法先进行点云数据点的离散高斯映射, 并将映射点集聚类; 然后使用自适应迭代过程得到两个或多个面的相交线上曲率值和法向量发生突变的尖锐特征点, 这些点与视角无关; 最后, 用改进的特征折线生长算法, 将特征点连接, 得到光顺特征线。实验证明, 该算法具有良好的自适应性、抗噪性和准确性, 是一种有效的三维模型特征线提取算法。

Algorithm for feature line extraction based on 3D point cloud models

This paper proposed a sharp feature line extraction algorithm of 3D point cloud models based on Gaussian map and curvature value analysis, which aimed to solve the problems that previous algorithms existed, including could not distinguish sharp and non-sharp feature points, the extracted feature points were relative to perspective, or feature points weren't connected. First, this algorithm conducted discrete Gaussian map for point cloud data, and clustered these mapping point sets. Then it used an adaptive iterative procedure to get sharp feature points, these points mainty located on the intersection line of two or more point cloud surfaces, where curvature value or normal vector suddenly changed and they were independent of perspective. Finally, it used an improved feature polyline propagation algorithm, connected the feature points, and got smoothing feature lines. Experiments show that the algorithm has good adaptability, noise immunity and accuracy, it is an effective feature extraction algorithm for 3D models.