Research on Image Matching in Printing Defects Detection Based on Machine Vision

Abstract

A new similarity algorithm based on image difference and template matching is proposed to solve the problem of real-time matching between the printing image and the template image based on the similarity algorithm in online printing detection. According to this similarity, besides judging whether the printing image matches with the template, whether the printing image is defective or not, it can also judge whether the printing image is offset or deflected relative to the template. If there is a deflection, Hough transform based on mathematical morphology is used to calculate the deflection angle of the printing image relative to the template image, and then the printing image is corrected according to the deflection angle. If there is an offset, the offset of the printing image relative to the template image is calculated according to the principle of phase consistency, and then the printing image is corrected according to the offset. At the end of this paper, the comparative experiments with the similarity-based image matching method are carried out. The experimental results have verified the superiority of the image matching based on the similarity which is proposed in this paper. In addition, the experiments on the offset and deflection of the printing image relative to template image have been respectively carried out. The results verify the feasibility and validity of the proposed method in the paper.