An approach for extracting curve profiles based on scanned point cloud

Abstract To control product quality, it is essential to conduct surface detection and curve detection. The existing approaches are sufficiently developed in detecting surfaces while are limited in detecting curves. It is essential to extract the curve profiles from the product at first. An approach of extracting curve profiles from scanned point cloud is proposed. Firstly, a slice set is constructed to extract section line points from scanned point cloud. Secondly, based on normal angle of adjacent points, three pattern vectors are defined to distinguish fold-line section line, small-arc section line and large-arc section line. Thirdly, three methods are proposed to identify the curve-profile points on the three kinds of section lines. Qualification experiments are conducted on seven circle profiles and a rectangular profile. The results show that extracted curve-profile points deviate from reference data within ±0.1mm with standard deviations smaller than 0.07mm.