Multi-features selection and extraction of structured light images of vision sensor for seam tracking

According to various types of images features of welding work pieces joints in vision sensor for seam tracking in welding robot, a method of features selection and extraction for structured light images is proposed. Images are fragmented with adaptive thresholds, and the centerlines of laser stripes are extracted. By the characters of three typical joints: the groove joints, the butt joints and the left or right lap joints, the points of center gravity of grooves, turning points and the points of discontinuity of stripes are selected as features to be extracted respectively. Test results of images of different type joints testify the efficiency of the method.