An adaptive process of reverse engineering from point clouds to CAD models
Jin Liu
Abstract
In the field of industrial design, most manufactured objects are designed using 3D CAD (Computer-Aided Design) model. However, the original design may sometimes be lost or not available. A reverse engineering approach is then required to generate a geometric CAD model from 3D point clouds obtained by scanning objects. In this paper, an adaptive process for automatic reconstruction of CAD model from point clouds is proposed. First, the approach extracts primitive shapes from point clouds by RANSAC algorithm, then it analyses deviations of points from the fitted primitive shapes by histograms. For point cloud patches segmented unreasonable, the approach updates parameters of segmentation according to the Gaussian noise and repeats the primitive shape detection process. After certain rounds of iteration, the approach can detect reasonable primitive shapes from point clouds. Then a rule for obtaining CAD models by primitive shape alignment is introduced. The proposed approach is evaluated on datasets provided by the AIM@SHAPE Shape Repository.