Research On The Defect Detection Technology Of The Screen Printed Matter

Printed matter as carrying information medium,is closely connected with people's life,but because of living and appreciation standards constantly rising, the needs of higher quality and more diversified printed matter become more and more. In printing, the following problems will inevitably happen, such as:scratch, ink, panel, more printed and omission because of the craft and environmental reasons. When these things happened, due to several subjective and objective factors, the traditional human check methods can not complete the detection tasks accurately, quickly in a long time. At present, with the machine vision technology development, the use of machine vision for automated image quality print detection has become feasible and practicable.This paper uses the automation detection system of screen printed matter defect as the application background,combined with the technology of machine vision, digital image processing and screen printing etc.and do some research in key technology of printed matter defection systems.Firstly paper introduces the composition and characteristics of the machine vision system. In basis of analysis of the system's design goals and challenges, paper gives the overall design proposal. Secondly, based on characteristics and defects of screen printed matter,and the analysis of noise type acquision,establish the defect detection methods and ideas.In image preprocessing, in order to filter noise existence in and out of system, the paper adopt a kind of improved median filter algorithm based on the traditional median filter; In template making process, using statistical average method to finish the establishment of standard template;In image registration process, in order to ensure the undetected image and standard templates accurately comparison, and through a detailed research of two common printed matter image registration methods, paper bring out a improved image registration algorithm base on LTS Hausdorff distancecorner and image profile. The algorithm is based on distance as similar to avoid establish accurate measure of the point-to-point relathionship,at the same time extracting image profile as registraion feature information,greatly reduced the amount of matching calculations. finally through calculating the centroid coordinates of two images to determine lateral-torsion parameters of inspected image,and calculating the image profile's law angle relation to acquire the inspected angle. Experimental results show that this algorithm has both the real-time and accuracy, and achieved good matching effect;In defects detection process, firstly analyzing three printed matter defects detection algorithm based on image processing technology, and then elaborate on their respective strengths and weakness, at last paper brings out the connectivity analysis algorithm based on chain and sequential,to get the shape,perimeter and area of defect information. and the algorithm resolve the problem of can not accurate statistician defect information.The paper use Visual C++6.0 as the development tool to developing the program of printed matter image preprocessing, template, registration and defects detection. Finally, we use different types of defects in five groups of printed matter as the experimental samples, for the further verification of all the algorithm mentioned above, and according to the result, we analysis the real-time and feasibility of actual application of this printed matter defects detection system.