The Method Research Of Detection Of Color Printing Defect

At present, with the rapid development of technology and economy, people has higher requirement for the appearance quality of print. But in the process of production, due to the influence of environment, print always have all kinds of defects, such as omit、stain、overprint deviation and so on. In domestic, most of printing enterprises to the defect inspection still stay in the artificial phase, and it not only consume much time but also cannot assure the quality. The technology of automatic inspection is still in the infancy, so the research of printing defect automatic inspection based on machine vision and image processing can improve the inspection level of China’s print industry, and it has important realistic significance.To settle the question of color printing overprint deviation, a new algorithm based on template matching is put forward here. Firstly, Canny’s edge detection method is used to get image edges, and the candidate region of color code is found by template matching the shapes along edge’s direction according to the similarity measure; Secondly, the color information of candidate region is used to get accurate position, and Gaussian pyramid decomposition and acceleration termination matching strategies are adopted to improve speed of the algorithm in the process of template matching. Experimental results show that precision of the algorithm in the automatic inspection of overprint deviation inspection is higher than overprint precision requirements of national printing industry standard, and it can meet the real-time requirement in the industrial printing inspection, and it can inspection more than twelve kinds of color, suitable for multicolor printing system.In order to inspection the defects of the print, precision registration between standard image and real image is a premise for print quality inspection. Whether we can implementation the registration between standard image and real image is a direct impact on the performance of inspection system. The features are invariant to image rotation, translation and scaling. So we proposed the image matching method based on SIFT algorithm. Firstly, image registration methods based on feature contexts extract feature information of the standard image and real image, generate the feature descriptor. Then, we need to establish the corresponding relationship between the standard image and the real image via the feature points. Lastly, accomplish the image registration according to the relationship between feature contents. Experimental results show that this algorithm has the character of stability, it is invariable in the scale and rotation in the image changes, it also has a strong adaptation in illumination and image deformation, and it can meet industrial requirement to the rapidity.In achieving the registration of the standard image and real image, we use method of the image differencing, image binarization, binary image morphological processing and the analysis of the connected components. In the end we can take the area as threshold to judge the printed image whether the qualified products, or make the system take waste instruction.