Abstract: In the current welding seam detection process of offshore platforms, image edge segmentation is achieved by relying on the Sobel operator, which is sensitive to noise information, resulting in lower ACC values (accuracy) of the welding seam detection results. Therefore, a welding seam detection method for offshore platform welded components based on Canny operator edge segmentation was proposed. Using the concepts of expansion and corrosion in mathematical morphology, the welding seam images of offshore platform welding components were filtered. The preprocessed images were input into the visual attention model, the comprehensive saliency features contained in the image were extracted, and the visual salient regions of the image were annotated. Based on the Canny operator edge segmentation algorithm, all edge pixels in the image were detected, and edge connections were achieved through dilation and refinement operations to segment the entire target area from the image. A seed region was selected and the region growth results were obtained according to the growth judgment criteria, in order to obtain the weld seam detection results. The experimental results showed that the ACC value of the new design method for weld seam detection was set above 0.95, which well met the safety inspection requirements of welded components on offshore platforms.