Research On Surface Defect Detection Technology Of Steel Pipe Based On Machine Vision

Surface quality of steel pipe is an important index to evaluate the grade of steelpipe, which is good or not also directly affects performance and quality of theproduct. But when the enterprise is in the process of the production of steel pipe,scratches, roller printing, cracks, inclusions, pits, abrasion and other different typesof defects will inevitably appear on the surface of steel tube because of theenvironment, equipment and other reasons. These flaws not only affect theappearance of products but also reduce the performance of product. At present, mostof steel production enterprises arrange for some specific persons to detect the surfacedefect of steel pipe online, that will not only affect the production efficiency, butalso when people still work in a long time, they become easily fatigue. Thenundetected probability of the products will increase, and the qualified rate ofproducts will decrease，the quality can not be guaranteed. Therefore, an on-linedetection system of defect is developed to improve the surface quality of theproducts in the enterprises.The author puts forward a scheme for defect detect of external surface of steeltube based on machine vision,after doing research on technology of surface defectdetect of steel tube in domestic and foreign. Then the pictures,which is collectedfrom field,is tested, and the experimental results show that the scheme is feasibleand effective. This thesis focuses on the following issues:1System design of surface defect detection of steel pipe based on machine vision,selection of hardware and building of system platform.2The author analyzes main defect types which influence the surface quality of thesteel pipe and the causes of the defects through the picture collected from the field.3Theoretical research on surface defect detection of steel tube and algorithm toachieve, including image preprocessing, image segmentation, defect location anddefect labeling. On image preprocessing, the author introduces a kind of improvedmedian filtering algorithm for image smoothing. Then putting forward a fusion algorithm of Canny and morphological gradient to separate defect from background.At last, using a unique denoising algorithm for removal of interference and thepseudo defects, and locating and marking defect.4Recognition and classification of surface defect of steel pipe, including featureextraction of image with defects and the design of classifier. BP neural networkalgorithm is used to design classifier of defect, which is used to train, classify andrecognize samples collected.