Research On Defect Detection Technology For Printed Matter Based On Machine Vision

At present,the pad-transfer printing plays an important role in electronics,toys and other decorative printing,but it always appears a variety of printing defects due to the printing process,mechanical precision and so on.Traditional methods of artificial detection rely on subjective experience,which is unstable and inefficient.With the development of science and technology,the defect detection technology based on machine vision emerges.However,there is a general contradiction between the accuracy and the error rate,this paper studies the defect detection technology based on machine vision,propose the defect detection algorithm,builds the experimental platform,and verifies the feasibility of this scheme by analyzing the experimental results.For the characteristics of pad-transfer printing process and the requirements of the project,this paper designes the overall solution and hardware architecture and the overall solution of the paper,including the detection process.Based on this,builds the experimental platform.The image has some inclination and deviation in space,therefore,image registration is the key to defect detection.After making the standard template image,image preprocessing,studying the classical image registration algorithm,a suitable image registration algorithm is proposed based on the actual situation of the thesis.According to the characteristics of the branch separation of text image,the method of Fourier transform and Hough transformation is used to correct the inclination angle,then the Fourier phase correlation principle is used to correct the deviation,and the image with the picture is corrected by using the three-template matching algorithm based on the surf algorithm.The defect detection algorithm is also an important part of this paper after the inspection and standard template are matched.After some typical defect defection algorithm are discuseed and analyzed,the image difference method based on Gaussian sliding window is proposed,furthermore,image difference method based on difference of Gaussian gradient and Gaussian sliding window is proposed to solve the contour residual difference.Then a threshold method is used to generate the difference image,the eight-neighborhood floodfill method is applied to identify and count the defect.Finally,the paper analyzes the data and the cause of this result,and verifies the validity of the algorithm.