3D vision-based weld seam extraction method

Abstract

Butt joints and angle joints exist in all types of industrial production and are often taught for mass production, while the efficiency of teaching is greatly reduced for small batches and frequent changes in the welding environment. Aiming at the problems of small-lot and other production, a 3D camera-based fast identification system for common weld seams is proposed, which is able to quickly identify the weld seam position information. In this paper, based on the 3D large field of view camera and welding robotic arm, first measure the actual accuracy of the 3D large field of view camera, and then calibrate the camera by hand and eye, the calibration error is within the usable range, and finally propose a rapid identification algorithm based on the point cloud weld to identify the weld starting point. Experimental verification shows that the position error is less than 2.79mm to meet the welding process requirements.