Robots have been more and more widely used in the factories nowadays. Arc welding is a very important application for industrial robots. However, how to eliminate the deviation caused by the inconsistent workpieces for robotic arc welding is always plagued operators. For solving this problem, researchers have developed different kinds of seam tracking sensors, such as touch sensor and arc sensor. Now, vision, as a widely developed sensor, is applied to be the seam tracking sensor....[ Read more ]

Robots have been more and more widely used in the factories nowadays. Arc welding is a very important application for industrial robots. However, how to eliminate the deviation caused by the inconsistent workpieces for robotic arc welding is always plagued operators. For solving this problem, researchers have developed different kinds of seam tracking sensors, such as touch sensor and arc sensor. Now, vision, as a widely developed sensor, is applied to be the seam tracking sensor.

In this thesis, a vision-based seam tracking system is developed. The tracking sensor includes a camera and a structured laser. Rigid body motion and solid geometry are the base of tracking theory. For tracker alignment, eye-in-hand method based on relative movement of end-effector connected to a robot is proposed. It determines first the rotation then the translation. In addition, a simply method of calibrating the structured laser plan is given out. In Cartesian space, the position values of each feature point on the weld seam can be obtained though combining the parameters of camera and laser plane.

Except the tracking algorithm, some image processing methods, such as mean-value, thinning and Hough transform, are used to improve the image quality and tracking accuracy. Finally, a series of experiments based on Motoman MA1400 arc welding robot are setup to verify the whole algorithm and the reliability of the tracking system. The tracking accuracy we get finally is up to ±0.5mm.