Welding seam detection and tracking based on laser vision for robotic arc welding

Abstract

The welding robots are widely employed in the manufacturing field. In order to measure up to the developing need of automation and intelligent production, real-time welding process sensing and controlling are necessary, in which the welding seam detection and tracking is representative task. Among the kinds of sensors applied in the robotic welding process, vision sensors is one of the most widely used for automatic welding seam tracking and quality control in robotic arc welding. This paper is to present a new framework and the data process algorithm of the sensor. A novel two-step calibration approach is proposed in this paper based on the “Calibration toolbox in Matlab”, the proposed approach acquires the laser light plane by project the linear stripe on the calibration pattern plate during camera calibration process. Experiment result indicates that the fitting error of the detected 3D point data of the weld groove is less than 0.1mm. The tracking error in robotic welding process is no more than 0.2mm. Which is acceptable for robotic welding system.