Abstract

To design a stable laser vision seam-tracking system, an advanced weld image processing algorithm based on Siamese networks is investigated and proposed to resist the interference of arc and spatter in the welding process. This specially designed neural network, combined with powerful feature expression capabilities of deep learning, takes two welding images with different sizes as inputs and generates a target confidence map in a single forward pass by using the cross-correlation algorithm. To prevent the error accumulation and model drift, an online update strategy via local cosine similarity is developed. The use of metal inert-gas welding can realize real-time and precious tracking under the condition that the strong arc continuously shields the welding seam feature points.

© 2018 Optical Society of America

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)

Weld feature extraction method based on inter-frame constraint and dynamic template

Zhen Mei, Lizhe Qi, Min Xu, and Yunquan Sun
J. Opt. Soc. Am. A 39(5) 771-781 (2022)

Optical emissions from weld arcs and their effects on the performance of welding robot vision systems

Alan L. Lenef and Chester S. Gardner
Appl. Opt. 24(16) 2587-2595 (1985)

Three-dimensional reconstruction of mobile binocular stereo vision based on push-broom line structured light for a workpiece surface

Yue Wang, Xueyou Han, Jing Rui, Hailan Zhang, Lei Yin, Xuefeng Zhang, and Xiangjun Wang
J. Opt. Soc. Am. A 40(11) 2096-2107 (2023)

Polarimetric learning: a Siamese approach to learning distance metrics of algal Mueller matrix images

Xianpeng Li, Ran Liao, Hui Ma, Priscilla T. Y. Leung, and Meng Yan
Appl. Opt. 57(14) 3829-3837 (2018)