Abstract: Due to the variable position of the fillet weld between the prismatic pipe and the flange, and the low machining precision of the end face of the prismatic pipe and the low degree of automated welding in actual production, a set of circular welding of the prismatic pipe and the flange is constructed in this paper. The automatic welding system of the seam has a great application value for improving the automation degree of the welding of the prismatic pipe and the flange. The system uses a CMOS camera and a single-stripe laser to form a laser vision sensor to obtain the position and gap information of the corner seam. According to the characteristics of the collected images and workpieces, a suitable image processing algorithm is designed. First, the methods of grayscale transformation, mean filtering and morphological processing are used to preprocess the images. Appropriate threshold value and extreme value method are used to extract the center point of the light stripe. Finally, the Hough transform is used to fit the straight line and extract the position information of the corner seam. The laser stripe endpoint search method is proposed to extract the size of the welding seam gap. The results show that the image processing scheme has good effect and strong anti-interference ability, and can accurately extract the center position and gap size of the weld, which can well meet the requirements of welding robots for welding seam tracking and realize automatic welding.Highlights: (1) Design an automatic welding system suitable for prismatic pipe and flange fillet welds.(2) Design an algorithm for extracting the center position of the weld seam for this system.(3) Design a corner seam gap extraction algorithm.