Research On Defect Detection Technology Of Color Card Based On Machine Vision

With more and more attention paid to the quality of products produced in China,OEM factories in the clothing industry have relevant requirements on the inspection of commodity signs.In the past,this kind of detection work was completed by artificial eye inspection,and its work efficiency and detection effect were not very good.These signs are usually colored cards,with a variety of types and complex image information.With the continuous development of computer hardware and software technology,automatic detection technology based on machine vision can solve these problems.In this paper,the appearance defect detection technology is studied in detail by taking the color card as the research object.The main research contents are as follows:1.Designed the overall scheme of the color card detection system according to the actual production testing requirements.Reasonable construction of the system components,and complete the functional design.The industrial camera,lens and light source are studied.2.According to the problems of low contrast and noise in image acquisition in practical detection,the corresponding solutions are studied,and appropriate image processing methods are selected through experimental comparison.Aiming at the problem that the matching template can accurately capture the position of the color card in the image the threshold segmentation method is studied and selected.3.Based on the research on the template matching principle,a defect detection method of color card is constructed to realize the defect detection function of color card.The designed detection algorithm is realized by writing detection software,and the rationality of the designed color card detection method is verified by experiments.4.Based on the research of convolutional neural network,a color card defect classification method is constructed to realize the classification of color card defect images.In the design of the network model,the network structure of VGG-11 is referred.Based on the design idea of improving the performance of convolutional network,Compared with the single 3×3 convolution kernel used in the network in VGG-11,two InceptionV2 modules are used in the back of the newly designed network,which is the combination of 3×3 and 1×1 convolution kernel,and the performance is better.The data set is made to train the network model,and the correctness of the design idea is verified through the experimental test,and the defect classification of the color card is realized.