关键词: 三维点云, 模糊推理, 法向量估值, 尖锐特征, 薄片特征

Abstract:

Before a three-dimension reconstruction,the normal vector should be estimated because it may be unreliabledue to the error of getting point clouds. In this paper,a fuzzy normal vector estimation algorithm for the objectwith any shape is proposed after analyzing the existing estimation algorithms. In this algorithm,first,the k-nearestneighbor value and the curvature of the cloud point data are input into a fuzzy inference system.Next,the pointclouds are classified according to fuzzy inference rules,and the parts with thin or sharp features are distinguishedfrom the point clouds of model and are then estimated with a checker and with the attachment point algorithm.Finally,the proposed algorithm is evaluated by using a denture model with several kinds of point clouds.The resultsshow that the algorithm is of high estimation accuracy,simplicity and feasibility.

Key words: three-dimension point cloud, fuzzy inference, normal vector estimation, sharp feature, thin feature