Technology And Method Of Printing Character Defects Detection Based On Machine Vision

With the development of the modern printing technology,product manuals,product packaging and other printing products has realized automatic printing.However,in the process of mass printing,various defects are inevitable,and of course character printing is no exception.Characters defects often appear in the process of printing,such as character omission,misprint,blot,misplacement,etc.It is of great significance to eliminate the products containing printing defects to improve the comprehensive quality and market competitiveness of the products.It is easy to cause visual fatigue and low detection efficiency through the method of artificial observation to pick out defects from dense characters.At present,the machine vision technology is booming in China,the development level of hardware and software is improved,and the defects detection based on machine vision is feasible.This paper studies the defects of printing characters and the methods of detecting character defects based on machine vision.Firstly,the scanner is used to acquire images because of unique advantages on flat scanning,avoiding the influence of character defect detection on camera distortion and light sensitivity.Secondly,by extracting Sift feature points of the template and test images,use the minimum Euclidean distance method to match these points.This method can find the matching feature points in the template image of each Sift feature point in the test image,and the two pairs of feature points are realized.Then from all of the matched filter three optimal matching and the affine transform algorithm with the three best matching Sift feature points is used to correct and locate.The coordinate space of the test image is transformed into the coordinate space of the template image to realize the positioning function of the target.It can improve the positioning accuracy by making full use of well-preserved Sift features in the defect images and positioning error is controlled within one pixel.The positioning method avoids the inaccurate positioning results from the defect characters,and the improvement of positioning accuracy lays a good foundation forthe identification of defects.After the localization of the test images,the threshold binarization segmentation is performed.Then the pixel difference method can be used to accurately separate the difference between test images and template images.In view of the characteristics of false defects,the segmentation defect image is filtered to eliminate false defects through the designed filter operator.Then,the defect is marked by the seed filling connected area marking algorithm,and the alarm information is given.C++ programming language and OpenCV are used to realize algorithm and an interface is written by using the QT language.This system can successful detect printing defects and complete the output of the prompt message.