Vision sensing and surface fitting for real-time detection of tight butt joints

Weld joint detecting technology is significant and indispensable to intelligent welding manufacturing. And vision sensing is considered as a promising weld joint detecting technology, with the advantages of rich information, non-contact measuring, high accuracy and anti-electromagnetic interference. Many structural components used in the aerospace field are manufactured through welding with tight butt joints. However, the conventional structured light method is not suitable for the detection of tight butt joints, since they rely on the light stripes' geometric distortion caused by the grooves. This study presents a real-time visual method for the tight butt joints' detection. Uniform surface light is employed to highlight tight butt joint and acquire 2D information, and structured light with a double-line pattern is projected onto the seam to acquire 3D point cloud data by optical trigonometry. Surface fitting with moving Least Squares (MLS) method is conducted in an iterative manner to combine 2D and 3D information and calculate the seam's positions and poses. Experimental results indicate that the proposed method is capable of detecting stably and accurately tight butt joints with a gap of less than 0.2 mm. The technique has the potential to be applied in tracking tight butt joints of complex structures in the aerospace industry for an automatic welding process.