The Research Of Detection System For Soft Trademark Printing Defect

The detection system based on digital image processing algorithm and machine vision technology could overcome the deficiencies of traditional manual detection methods, such as the low efficiency and accurate, unstable result and the waste of manpower and material resources, and meet the practical application requirement of enterprise and improve the production efficiency and economic benefits.This paper realized the design of integral structure of on-line detection system according to the production status of a soft trademark enterprise and combined the performance of whole system, processor structure, hardware system and software system, and completed the image matching, printing defect extraction and defect pattern recognition by the way of digital image processing.During pretreatment process, we got the standard template image and matching template image by statistical average method and optimized the corresponding size parameter and coordinate parameter, and realized color space conversion and gray processing of color image.The space searching algorithm is optimized by genetic algorithm and measured image is matched with template image with matching features extraction, similarity parameter selection and searching space optimization, and the parameters of GA are simulated and improved during image matching process.Image binarization processing is been done and the features,such as defect quantity, defect region and corresponding area, edge, centroid, external matrix, are extracted in defect detection process and the the features matrix is been generated.We made the different feature as the input of algorithm and then constructed the feature matrix models, and completed the defect recognition and performance analysis by PNN, GRNN and SVM, realized the defect pattern recognition of soft printing trademark.