Abstract: Aimed at the impact of various factors such as arc light,welding wire,and arc occlusion during pipeline welding,which made it difficult to accurately extract the edge of the molten pool contour,an improved traditional Canny operator-based molten pool edge detection algorithm was proposed. Firstly,guided filtering was used to replace Gaussian filtering for edge preservation and denoising of the molten pool image. Secondly,gradient templates in the 45° and 135° diagonal directions were added to the horizontal and vertical directions for detection,and gradient components were calculated. Finally,local adaptive thresholds were used to replace the global fixed method of dual thresholds in non-maximum suppression techniques to complete the edge detection of the molten pool.At the same time,morphological processing techniques were combined to improve the noise resistance and adaptability of edge extraction. The experimental results show that through different swing width tests,the extracted molten pool width through the improved algorithm is being verified with the weld seam width after welding. The accuracy of the extracted molten pool width is greater than 95.56 %,and the error does not exceed 0.3 mm,which meets the accuracy requirements of GMAW molten pool width measurement.

Key words: pipeline welding, Canny operator, edge extraction, morphological processing