Quality Inspection And Fault Analysis Of Mental Can Printing

The quality inspection mainly depends on manual or instrument in the process of mental can printing,and its accuracy can not meet the needs of industrial production.In order to meet the needs of modern production of mental cans,experiment uses the method of combining machine vision and image processing to carry out feature extraction and classification detection of metal cans printing defects,and realizes machine vision detection to meet the needs of enterprises for high-quality metal cans printing products.The methods of image processing and data statistics are used to realize the extraction,classification and geometric feature analysis of the printing defect image of metal cans,and the system platform of defect detection is established in the software environment of visual studio C++and OpenCV.Firstly,the image is preprocessed,including graying and noise removal.Secondly,the foreground of defect part is separated.SIFT algorithm is used to match the standard image and the image to be detected,and grabcut algorithm is used to segment the target region in foreground separation.Thirdly,six texture features based on gray level co-occurrence matrix and ten geometric features are used to train and classify the defects of metal can printing by BP neural network.Finally,Finally,the location and quantity of defects are extracted based on the geometric contour features of defects.The results show that:Median filter can reduce noise of image;SIFT algorithm is used to extract and match image feature points effectively;Image difference and grabcut segmentation algorithm are used to separate the foreground of defects;BP neural network training is used to classify defects;The outline of the defect is marked.According to the location and number of defects,the defect contour is marked.