Research On Off-line Inspection System Of Printing Overprint Accuracy Based On Machine Vision

The printing industry is not only an important carrier of the country's cultural soft power,but also an indispensable part of the country's real economy.In recent years,with the increasing improvement of people's living standards,people's requirements on the quality and beauty of printed matter have been improved,and overprint accuracy plays a crucial role in the quality of printed matter.The online printer register control system can detect the register accuracy in real time during the printing process,so as to adjust the corresponding equipment to improve the printing quality.However,for third-party inspection agencies,when inspecting print quality,off-line registration accuracy inspection systems are often required to ensure the stability and reliability of the testing results.Under this background,the research and experiment on the algorithm model of overprinting accuracy detection of color printed matter based on the machine vision system are carried out.The main tasks are as follows:First of all,based on the theoretical knowledge of digital image processing,the overprinting logo images of color prints are preprocessed.Unify the digitized pictures of the sample to be tested to the specified size,and fix the dimension of the pixel matrix;convert the color overprint logo image to the CMYK space and then segment the image based on black;use 3*3 median filter and 3*3 Gaussian filter combined filtering algorithm denoising and equalizing,removing interference and useless information,making the useful information of the image more prominent;using the Canny edge detection operator to extract the image edge information in the feature generation process,for the calibration and prediction of the overprint precision detection model,Which is beneficial to improve the speed and precision of the overprint detection algorithm.Secondly,on the basis of the pre-processing link,an overprint accuracy detection algorithm based on straight line slope detection and two overprint accuracy detection algorithms based on neural network are proposed.The detection method based on the slope of the straight line detects the slope of the overprint line of each color,and at the same time,the unit pixel and the actual distance are calibrated to detect the overprint accuracy.However,the experiments show that this method requires high image edge detection.On this basis,the model training was carried out based on the full connected neural network and the convolutional neural network,and the overprinting precision image data that has been input to complete the detection is used as a label for model training,and multiple sets of comparative experiments are used to optimize the hyperparameters and determine the parameters.Compared with the linear slope detection algorithm,the neural network model algorithm is more versatile,the model has good adaptability,and it truly solves the subjectivity of people and improves the detection accuracy.Finally,the hardware visual inspection experimental platform was built,and the GUI functional design based on MATLAB was used to realize the intuitive operation interface,including the image preprocessing part and the overprint accuracy detection algorithm part.The software reliability test proves that the visual inspection system runs stably,and the detected data is reliable and efficient.