Defect Detection Online For Printed Matter Based On SVM

The printing is closely related with people’s daily lives, with the rapid development of digital networks, although the printing industry has been some impact, but still need a variety of printed matter in people’s lives. Traditional printing defect detection technology has been unable to meet the demand for speed and accuracy of the printing industry, printing quality assurance need to print defect detection technology with the times. In recent years, the printing defect detection technology based on machine vision gradually been brought by, however, the development since the actual detection result is not satisfactory, machine vision, based on statistical learning theory, in-depth study of the print defect detection technology, the main content divided into the following section:(1) The proposed selection method based on matching the promoter region of the edge density, high-speed parallel computing based on the CUDA the image cross-correlation matching algorithm, and matching accuracy were analyzed.(2) Based on directional features and differential search method based on the homogeneity of the neighborhood, presents a method based the the DOG gradient and the neighborhood of differential search construct difference images.(3) Select the difference image of the OTSU segmentation value and the ROI maximum light intensity of stimulation as compared characteristics. Then use the positive and negative samples of learning based on SVM training, and finally determine the judgment to be the test samples for defects by SVM classification.