Abstract

Welding as an industrial tailor is key to intelligent manufacturing technology. With the development of robots and digital manufacturing technology, autonomous programming intelligent welding robots have emerged. The seam tracking system based on active vision is widely used because of its high precision and obvious characteristics. However, the complexity of the welding environment, the interference of arc splash, and the reflection and local deformation of the workpiece will affect the extraction of weld characteristics. Based on the correlation and difference of adjacent frames in the continuous welding process, this paper combines the point and line characteristics of laser lines and proposes a weld feature extraction method based on an inter-frame constraint and dynamic template. The experimental results show that the distance error of the proposed method is within three pixels, and the average processing time of each frame is 9.6 ms. This method can realize weld feature extraction in the case of object reflection, arc splash, and local deformation, and has certain robustness.

© 2022 Optica Publishing Group

Learning Siamese networks for laser vision seam tracking

Yanbiao Zou, Jinchao Li, Xiangzhi Chen, and Rui Lan
J. Opt. Soc. Am. A 35(11) 1805-1813 (2018)

Center extraction method for reflected metallic surface fringes based on line structured light

Limei Song, Jinsheng He, and Yunpeng Li
J. Opt. Soc. Am. A 41(3) 550-559 (2024)

Robust seam tracking via a deep learning framework combining tracking and detection

Yanbiao Zou, Rui Lan, Xianzhong Wei, and Jiaxin Chen
Appl. Opt. 59(14) 4321-4331 (2020)

Keyhole morphology monitoring of laser welding based on image processing and principal component analysis

Ting Lei, Shiyang Gu, and Huiwen Yu
Appl. Opt. 61(6) 1492-1499 (2022)

3D pose estimation of large and complicated workpieces based on binocular stereo vision

Zhifeng Luo, Ke Zhang, Zhigang Wang, Jian Zheng, and Yixin Chen
Appl. Opt. 56(24) 6822-6836 (2017)