Abstract:Vision sensor systems with an auxiliary light source such as structured light–based vision sensor have been widely used for weld seam detection. However, the main drawback of this method is the preview distance between the welding position and the sensing position, which can generate unavoidable detecting errors. Laser welding of the small workpiece or narrow butt joint is especially vulnerable to detection error caused by preview distance. Meanwhile, only one point of the weld seam can be measured each time, which makes the corresponding image processing very sensitive to light noise. Consequently, a seam measurement method based on a passive vision sensor for narrow gap butt joint is proposed in this article. The weld pool is observed directly by this vision sensor so that there is no preview distance. An adequately adjusted telecentric lens is used to increase the contrast between the seam feature and the background welding noise. By this manner, a long and noticeable weld seam feature, as well as the weld pool, can be obtained in the captured images. A corresponding image processing algorithm based on particle filter is designed to extract the captured long weld seam. The particle filter is used to track the slope and intercept of the weld seam, which combines the advantage of both particle filter and Hough transform. As the particle filter algorithm could generate the result by evaluating both the previous and the current measurement results, the corresponding image processing can be robust even when a few images are of poor quality. Finally, laser welding experiments were carried out, and the results revealed that the proposed method could achieve detection accuracy of 0.08 mm when welding 0.1 mm width narrow butt joint at 2000 mm/min welding speed.

Robust weld seam tracking method based on detection and tracking of laser stripe

LSFP-tracker: an autonomous laser stripe feature point extraction algorithm based on siamese network for robotic welding seam tracking

An automatic feature point extraction method based on laser vision for robotic multi-layer multi-pass weld seam tracking

Additive seam tracking technology based on laser vision

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

Feature Extraction of Welding Seam Image Based on Laser Vision

Active visual continuous seam tracking based on adaptive feature detection and particle filter tracking

Weld seam track identification for industrial robot based on illumination correction and center point extraction

Real-time detection and localization method for weld seam of narrow butt joint based on semantic segmentation

A Robust Weld Seam Tracking Method by Using Intensity-Modulated Structured Light

Robust welding seam tracking using image seam extraction

Weld Seam Recognition Based On CCD Image Processing

A Feature Matching Algorithm Applied to Welding Trajectory Revising

Research on a Feature Point Detection Algorithm for Weld Images Based on Deep Learning

Welding seam recognition and tracking for a novel mobile welding robot based on multi-layer sensing strategy

A Novel Seam Tracking Technique with a Four-Step Method and Experimental Investigation of Robotic Welding Oriented to Complex Welding Seam

A robust weld seam detection method based on particle filter for laser welding by using a passive vision sensor

A robust butt welding seam finding technique for intelligent robotic welding system using active laser vision

A real-time seam tracking system based on extraction and restoration model in robotic GMAW

Research on Trajectory Recognition and Control Technology of Real-Time Tracking Welding

Robotic Seam Tracking System Combining Lightweight Segmentation Network Design and ADMM-Based Structured Pruning