A visual recognition algorithms for seam feature extraction of perforated plug welding

A visual recognition algorithms for seam feature extraction was studied for intelligent adaptive welding of the perforated plug joint between steel-plate and rebar in the shield building of the third generation nuclear power station. In view of the complex characteristics on plug hole size, root gap, inclination angle of perforated plug joint, the double stripe laser vision sensing technology based on the discrete data simulation relation curve was used to identify the area to be welded before welding, The feature datas affecting weld formation were extracted, including the center coordinates and diameter of plug welding hole, the height of rebar-end above steel-plate, the inclination angle of joint, the gap of weld seam, etc., which provide the basic data for the joint planning of welding torch’s gesture and welding process parameters. The error analysis of the extraction accuracy of the algorithm was carried out by comparing the measured sample data. The results show that the center coordinates deviation in the x, y direction are within the range of ± 0.2 mm, the angle deviation in the R_x, R_y direction is within the range of ± 0.2 degree, which realizes the high precision and reliability of feature extraction of perforated plug welding.