Online Detection System Of Printed Surface Quality Of Medical Packaging Based On Machine Vision

With the development of printing technology,the speed of printing equipment has been increased to more than 100 meters per minute at present.Printing accuracy is as high as submillimeter,which is a challenge to printing detection system.On account of machine vision,the continuous online detection system can overcome shortcomings from traditional manual inspection of printing quality detection such as missing detection,detection of inconsistent standards,difficult to fulfill tasks.Automatic,efficient,accurate and reliable are the features of this system,which meets the requirements of the rapid development of the industrial automation under the national background.Based on the relevant technologies of domestic and foreign,the purpose of this study is to improve the detection accuracy and detection speed of printing defects automatic detection,to design and accomplish the machine vision based online detection system of pharmaceutical packaging printing quality from the printing image acquisition,image processing algorithm and the parallel acceleration and other aspects of the algorithm respectively.The image of the printed matter is the prerequisite of determining the accuracy of the detection system of printed matter.Originally,this study analyzes the camera sensor,lens,shooting accuracy,shooting speed,light source and lighting method,uses 4K linear CCD high speed industrial camera and LED White strip light source and other major hardware devices to design the image acquisition system to ensure the system to obtain reliable,high-definition print image;Secondly,from the shooting principle and spatial coordinate system and other aspects of the line camera scanning process of image distortion in detail.In this study,we propose a method for correcting the lateral misplaced distortion of the linear array camera in poor shooting environment,and based on OpenCV and CUDA programming.The results of this study show that this system can be used in high-speed printing production line to obtain clear and stable large-width real-time print images.For the print image processing part,first apply the GPU parallel accelerat ion noise reduction to the collected image;Secondly,this paper proposes a small template rapid image registration to improve the real-time registration of the system,and use the qualified print template image and the difference principle of the printed image to be tested to detect the inconsistency between the two to find the defect;On the basis of image difference method and morphological method,the core part is proposed to remove the image difference pseudo contour.The BLOB defect analysis method is combined with different difference methods of image flat area and gradient region.The difference between the telescopic deformation difference and the noise difference is greatly removed in the difference image including the telescopic distortion,the noise difference and the true defect difference,and the true defect is retained at the same time.Finally,the program is realized by OpenCV and CUDA.The results show that the algorithm has little effect on the real defect while removing the pseudo defect,and the algorithm can accelerate the efficiency in parallel and ensure the accuracy and real-time of the detection system.This paper designs and realizes the online inspection system of medical packaging print quality based on machine vision,which greatly improves the detection precision and detection speed to meet the production demand of printing enterprises for automatic online inspection of printed matter.