3D Reconstruction of Weld Pool Surface in Pulsed GMAW by Passive Biprism Stereo Vision
Zhimin Liang, Hexi Chang, Qiyue Wang, Dianlong Wang, YuMing Zhang
Abstract
Geometrical information of the 3D weld pool surface provides clues to determine the penetration state whose monitoring and feedback control are crucial for critical applications involving high temperatures and pressures. In this letter, a biprism stereo vision system is established to sense the weld pool surface under different penetration states during pulsed gas metal arc welding (GMAW-P) with a V groove joint. Only one optical filter is used to block out the arc and capture a clear image of the weld pool during the base current period in GMAW-P. While most of the weld pool surface is textureless, there are a few features such as slag that are detectable. Hence, during the stereo matching sparse feature corners are detected and credibly matched in pairs, followed by a disparity region growing step to produce a semi-dense and unambiguous disparity map. Polynomial interpolation is, then, applied to obtain subpixel disparity for every matched pixel. Finally, the point cloud and triangle mesh representation of the weld pool surface with different penetration states, partial, full, and over penetration, are presented. The experimental results verified the effectiveness of the proposed biprism stereo vision method in monitoring the weld penetration state.
BibTeX
@inproceedings{ral2019_3dreconstruction,
title = {3D Reconstruction of Weld Pool Surface in Pulsed GMAW by Passive Biprism Stereo Vision},
author = {Zhimin Liang and Hexi Chang and Qiyue Wang and Dianlong Wang and YuMing Zhang},
booktitle = {RA-L 2019},
year = {2019}
}