Abstract: Since the traditional manual method to detect the scratches of small size pipe is complicated and may produce big human errors, a computer vision technology is adopted for the analysis and recognition of the scratches on the small size pipe surface, which uses the contour extraction of binary image to get the feature parameters of the scratches and analyzes them based on morphology. This method enables the features parameters, such as the area and the shortlong diameter ratio, to be extracted for identifying the scratches. It can effectively overcome the shortcomings of the traditional method, improve the stability and accuracy of the detection.