A V-shaped weld seam measuring system for large workpieces based on image recognition

First of all, it is necessary to accurately measure the spatial positions of weld grooves when the welding robot performs the automated welding of large workpieces. However, laser stripes are easily affected by light sources and environmental noises in the measurement process based on image recognition. The work designed an automatic measuring system for V-shaped weld seams of large workpieces. First, a mathematical model for calculating the spatial positions of welds was established, and then, a line laser was incident on workpiece grooves to form multiple line segments. Then, image analysis was performed on the intersection of each line segment. These intersection points were brought into the mathematical model after they were mapped to the feature points in the image to calculate the spatial position corresponding to each feature point. Finally, the spatial positions of all discrete points at different positions of the welding seam of the large workpiece were obtained by scanning multiple laser lines to complete the measurement of the entire welding seam. The work proposed a new idea of extracting feature points from weld images that combined Radon transformation, Harris corner detection, and mean clustering to improve the stability and automation of weld location extraction. Multiple groups of different parameter experiments were used to measure two groups of different bevel angles with the length, width, and height of 180 x 60 x 30 mm and 180 x 36.64 x 30 mm, respectively. The maximum measurement error was 0.653 mm, with a minimum measurement error of 0.001 mm, an average measurement error of 0.147 mm, and an average measurement accuracy above 99%. Results showed that the research method with high robustness and noise resistance rapidly could realize the automatic and accurate measurement of weld seams.