A method and device for detecting defects in printed matter based on artifact removal

Abstract

The invention discloses a printed product defect detection method and device based on artifact removal, the method comprising: image positioning and registration: based on the linemod feature point positioning and registration algorithm, the standard template image and the target image of the printed product to be detected are positioned and registered ; Target image artifact removal: Divide the standard image and the target image after positioning and registration into several sub-blocks of the same size, and use the sub-block neighborhood sliding artifact removal method to remove the artifacts caused by the local deformation of the target image; extract the final Defects in the background area of ​​the difference map: extract the contour of the standard image and perform mathematical morphological expansion operations to obtain the contour mask, use the contour mask covering method to segment the final difference map into contour and non-contour areas, extract and distinguish the non-contour areas of the final difference map And the defects in the contour area, and finally integrate the defects and output the display. The method and device of the invention can successfully detect defects such as dirt, incompleteness, double image, displacement, scratch, missing print and the like.

Description

technical field [0001] The invention belongs to the field of defect detection of industrial visual prints, and in particular relates to a method and device for detecting defects of prints based on artifact elimination. Background technique [0002] The types of surface defects of printed matter include dirt, incompleteness, ghosting, shifting, scratches, missing printing and color distortion. At present, manual visual inspection is generally used, and experienced workers use naked eye inspection, but long-term naked eye inspection will cause visual fatigue, which is likely to cause missed inspections or misjudgments. In addition, manual visual inspection is inefficient and expensive. In recent years, some detection technologies using image processing and analysis through machine vision have also appeared, but there are also various problems. The technical difficulties in the detection of printed matter defects include: compared with the standard template image, the image col...

AI Technical Summary

Problems solved by technology

[0003] The traditional method of removing artifacts based on morphological corrosion, when the image is corroded, the algorithm has no ability to distinguish between true and false defects. While removing false contours, some real defects with a relatively small area will also be removed, which affects the effect of printed matter detection.

In addition, the erosion and expansion of image morphology will change some features of the real defect points, such as the defect point area, centroid, shape, etc. These changes have a great adverse effect on the classification and recognition of subsequent defect points.

When the small-sample image defect detection method based on deep learning is applied to an industrial production line, there are very few products with surface defects in printed matter, and it takes time and effort to collect a large amount of defect data and pictures.

In addition, due to the variety of defects in paper prints, it is an unbounded problem; especially for the defects of missing printing, it is difficult to manually label the samples used to train the deep learning model

Images