A feature matching algorithm applied to welding trajectory revising

Laser welding robots usually adopt the method of “teach and playback”. Practically, if the weldment is placed inaccurately, it will have to be taught again. In order to solve this problem, we establish a high efficient system to revise welding trajectory automatically, which adopt improved SIFT algorithm to get the mapping relationship between standard weldments and on-line weldments. Firstly, we calibrate camera parameters based on planar constraint. Secondly, we adopt the improved SIFT algorithm to calculate the feature points, which can raise matching ratio and decrease the computing complexity. Next, we match key points of standard weldments and on-line weldments through kNN algorithm. Thirdly, we use the homography matrix to map the matched points from image coordinate to world coordinate so as to get affine matrix. Finally, we can achieve the goal of translating the welding trajectory between standard weldment and on-line weldment.