Abstract： Aimed at welding seam tracking of multilayer single pass welding by narrow gap rotating arc MAG, a passive vision sensing based welding deviation detection algorithm was proposed. First，the welding image was obtained by the welding position sensor and the CCD camera in trigger mode when the rotating arc was on the left or right side of the groove. Second, two adaptive thresholds were computed according to the grayscale histogram of the welding image. The large threshold was used to obtain the arc region and the arc central location. The small threshold was used to obtain the base metal area of the groove, and the edges of the groove were obtained by calculating the horizontal direction firstorder difference in special horizontal positions. Finally, by comparing the image process result of two images in each rotation arc cycle, the welding deviation, i.e. the difference between the center of arc rotating and the center of the groove, were obtained. The deviation detection algorithm is efficient and reliable. It can avoid the interference of the variation of the groove bottom, and be used for comparison or integration with other deviation detection algorithms.