Research On The Key Technology Of Defect Inspection Of Print

Image registration and image segmentation play decisive role in the defect inspection of printed material, and become a key factor in the success of the machine vision software design and development.Image registration technique: aiming at the printed texture feature and geometrical characteristics, this paper uses image registration algorithm of VE4000 vision system to realize, using template matching technique to design single template, and then according to the characteristics of images, cutting single template to splice registration image, and doing registration operation. Because the VE4000 forms registration image according to each image, it need to run a long time. On the basis of this, design a new template again, seamlessly splice seven templates to standard images, and through template matching, cut out registration image of the same size of the processing image in the standard image, finally do the subtraction operation for registration images of the two algorithms.Image segmentation technique: aiming at region features in prints, this paper respectively uses level set algorithm of C-V model and BLOB analysis technology for image segmentation. Due to image shooting angles, use pre processing algorithm, then make level set operation for images after preprocessing. Because of particularity of image features and algorithms, it failed to achieve the expected effect of image segmentation. Therefore, make further analysis of characteristics of image areas, use the BLOB analysis to do two value processing for image. BLOB analysis uses projection correction and bilinear interpolation algorithm, which can eventually get better segmentation effect for image.Based on the above image registration and image segmentation theory, make experiments for the corresponding prints. For the image registration algorithm in VE4000 and the improved operation, the two methods both can obtain registration image, then make subtraction operation for the two registration images to get the difference image. According to difference image, it can prove that the improved algorithm is better than the VE4000 registration algorithm for registration. Image segmentation is aiming at the nameplate printing, firstly use the C-V model in level set to make image segmentation, this algorithm did not meet the requirements of the segmentation effect; then analyze characteristics of image, use BLOB image analysis algorithm to do segmentation, this algorithm not only achieves image segmentation, but also has higher segmentation level, faster speed, stronger robustness, and fully meets the needs of practice.