Research On Printing Surface Defect Detection Based On Machine Vision

With the development and progress of social economy,the social demand for the apparent quality of printed matter has entered an almost harsh level.Aiming at printing defects such as misregister,leak printing,scratches and spot etc,it has become an inevitable trend and choice of industry development that by using scientific and accurate rapid detection and print quality detection(especially surface defects detection)based on machine vision.And considering above problems,a series of research and technology development of automatic identification and detection methods for the print typical defects have been done,which based on the industry background,facing the future application and supported by the theory and technology of digital image processing.The main works and achievements of this paper are as follows:(1)Firstly,this paper researches the detection technology of the print registration deviation based on the theory and method of image processing.This paper selects the traditional cross-line identifier and the improved four-color dot identifier as the research obj ects and proposes a new method for extracting four-color dots from printed images based on K-means clustering and floodfill algorithm,and the method improves the calculation efficiency of the deviation center and the calculation accuracy of detection.In order to overcome the information error caused by the limitation of the identifier set area,this paper improves the existing print registration detection algorithms based on print's own picture information,and proposes and verifies the segmentation method based on the statistical cumulative sorting.In the end,it turns out that the results of this algorithm comes closer manual precise measurements by comparing the results of manual precise measurement,pre-improved algorithm and improved algorithm,and in order to obtain this result,60 print images with print registration deviation were used.(2)Secondly,with regard to visual scanning image with defect detection for leak printing,scratches and spot,this paper researches general algorithm for shape defect detection based on image processing.On the one hand,using SIFT corner matching avoids the process of prefabricating template in template matching method,which guarantees the generality of the algorithm.On the other hand,the method which combining gamma transform with local dynamic threshold segmentation solves the problem of uneven image segmentation caused by uneven illumination.And based on that,this paper researches the print defects classification algorithms of scratches and spot based on CART decision tree.Finally using 200 images with scratches or spot defects as training set to construct decision tree,40 images as test set,and the classification accuracy rate is 93.3%.(3)Based on development environment and tools of VS?Qt?C++and OpenCV,a prototype system of surface defect detection based on machine vision is designed and developed.And the system includes the detection algorithm proposed in this paper,and realizes the basic functions of image selection,parameter setting,detection and analysis,defect classification and so on.In addition,according to users' own ideas and interests,the system can also help them define the defect detection process and run system step by step and display the image effect after one-step processing in real time.So,the system can be used as a learning platform to learn relevant knowledge,and verify the feasibility of the defect detection algorithm.