Determination of the Initial Welding Point for Multi-pass Welding Based on Laser Vision

Abstract

Most of the welding robots in practice are still teaching and playback mode. They are lack of the function to perceive the external environment, so it is difficult to carry out adaptive welding pass arrangement and position planning for medium thick plate. In this research, extraction procedure for feature points in laser stripe is presented. We get the weld groove profile with laser scanning. After processing the laser stripe with some image process methods, we get the two lowest feature points of the laser stripe including binarize image, medium filter for image, extracting the center of laser stripe. Then, two target feature points are extracted with linear fitting and distance formula. Finally, we convert the image coordinate value to the TCP coordinate of the robot according to the calibration matrixes. As a result, we restored the physical cross section of the weld bead, and can decide the initial welding position with the two feature points for every welding pass in the medium thick plates.