Design And Implementation Of Visual Inspection System For Packaging Inkjet Printing Defects

There is no product specific production information on the pre-printed product packaging film,so it is necessary to print label on the production line of product package.The printed labels include the date of production,place of production,and manufacturer.Due to the influence of human factors or abnormal conditions of the equipment,there may be quality problems such as printing leakage and printing errors,which affect the qualification rate of the product.The traditional inkjet printing defect detection is mainly manual detection,its accuracy rate is affected by subjective factors,and it brings high labor costs.With the continuous development of image processing and computer technology,achieving intelligent quality inspection by applying machine vision has become a trend.This thesis studies the key technologies of visual inspection for inkjet printing defects and completed the development of vision inspection software system.The main work of the thesis includes:Firstly,in view of the problem of pre-printed text and pattern interference in image registration,we proposed an image registration algorithm based on feature matching and phase correlation.The algorithm is a coarse-to-fine registration scheme.The feature matching is used to complete the rough registration of the image,and the phase correlation algorithm is used for fine registration.Experiments show that the image registration algorithm has good robustness and accuracy.Secondly,in order to solve the problem that the traditional algorithm is invalid or low accuracy due to the character tilt,character adhesion,background overlay and other conditions in the process of printing text labels,we proposed an inkjet printing defect detection algorithm based on convolutional neural network.The algorithm uses the convolutional neural network in the character sequence recognition network to extract the text line image features,calculates the visual similarity between each sub-region feature of the image and the anomaly-free sub-region feature dictionary to evaluate the abnormality of the image.Experiments show that the algorithm has good performance in different industrial printing scenarios.Third,based on the above algorithms,we designed and implemented a visual inspection system for inkjet printing defects.The system uses a layered and modular design method to achieve componentized development,and uses a message-driven design to reduce the coupling between modules.Aiming at the problem of image longitudinal stripes caused by the difference in light sensitivity between CIS pixels,an image linear compensation module is designed to improve the quality of images.Knapp-Kushner experiment shows that the system is better than the manual detection.The visual inspection system for inkjet printing defects has been applied in the packaging line of a food production enterprise.The accuracy of the inkjet printing defect detection has reached the standard of the relevant enterprise,which meets the actual industrial application needs and improves the production efficiency of the enterprise.