Design and Calibration of a Binocular Vision System for Arc Welding Robot

摘要

Arc welding is a complex process. In robot arc welding system, the welding quality is largely dependant on the good tuning of the mechanical device, welding data,etc. But this is a time-consuming work; this paper first proposes a new structure of binocular vision system for robotic welding, which can be used to improve the welding efficiency by leading initial welding position,welding seam tracking and real-time monitoring of dynamic welding pool. This visual sensor-based system works in passive way, and doesn't require active light source. The binocular sensors are fixed on the end-effector of welding robot, which is also the so-called eye-in-hand system. Further more, two mirrors are built in to make up of the reflection light path system, which delivers the weld image to CCD cameras. In order to accurately lead the torch to the initial weld position and track welding seam, Zhang Zhengyou's method has been applied to calibrate the two CCD cameras individually.The intrinsic and extrinsic parameters of the camera can be calculated from the feature points of the calibration board. The relative position and orientation of the two CCD cameras are then obtained through the extrinsic parameters of the two CCD cameras. Finally, the hand-eye relationship can be derived by several hand-eye movement and the obtained camera parameters. All the calibration tests have been conducted by using OpenCV.Experimental results show that the proposed method is reliable and practical for real applications.