|
Welding process automation and intellectualization has become the future develop trend in welding technology field. Currently the robotic welding process has been widely used in automotive, heavy industry, aerospace and many other fields. However, most of the robots today are teaching-playback robots; when facing with complex and unstable work environments, they cannot properly complete the task. The robot intellectualization technologies are that using the multiple sensors to adapt unknown, complex and changing environments. In this paper, we focus on the robot weld seam recognition and path planning research based on single vision sensor, which is one of the robot intellectualization technologies and can replace robot teaching process.Firstly, the hardware system is built, including the robot body, the robot control cabinet, industrial personal computer (IPC) and visual sensing system, in which the IPC and robot communicate by using the DeviceNet field bus. A compact and multi-function robot welding vision sensor is designed, not only greatly increasing the gun’s flexibility and accessibility, but completing multiple welding functions.Secondly, the vision sensing system is calibrated by using a modified Matlab toolbox to obtain the camera intrinsic parameters matrix, distortion parameters, and hand-eye matrix, etc. Compared to the traditional calibration methods, this method is fast, simple and reliable.Thirdly, a new straight seam recognition method, using the image morphology, is explored. We acquire the straight seam image using CCD camera, Then process the image using image morphology and Hough transformation, finally get the welding start point, seam direction and welding end point. The whole method is very intuitionistic and fast, with high self-adaptability and stronger robustness.Fourthly, The "one eye-double position" mode is applied to complete the stereo reconstruction of feature points. With robot tool coordinate transformation, system simplifies the calculation process into an ideal binocular stereo vision model, and obtains the positions of welding start point for, transition points and welding end point in world coordinate system, then fits into a three-dimensional straight line. The whole stereo reconstruction process is of high accuracy and stability.Finally, the actual experiments are conducted to optimize the algorithm and improve overall system’s accuracy and stability, also explore the best work range of the sensor and influence of parallax and camera distortion to the results. The results are that, after correcting, the absolute error in x, y direction is within 1mm, while z direction of 1.5mm, and the mean square error is ? x ? 0. 399,?y?0.533,?z?0.707, showing the system is of high accuracy and stability. At the same time, the ideal scope of sensor is between height 70mm and 130mm, as well as parallax and distortion have no real impact to the system. In a word, this system has a clear advantage for straight seam’s recognition and path planning at a close range, with of good application potential.
|