We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Real-time Defect Detection Method for Printed Images Based on Grayscale and Gradient Differences.
- Authors
Wang Yangping; Xu Shaowei; Zhu Zhengping; Sun Yue; Zhang Zhenghai
- Abstract
In a real-time quality inspection of printed matter based on machine vision, artifacts are induced by commonly used image difference methods, making the identification of defects difficult. Thus, to eliminate artifacts and improve detection rate of printing defects, this study proposed a method that combines grayscale and gradient differences. First, the grayscale difference between template image and inspected image was performed to determine the defect in the nonedge region according to the grayscale difference threshold of non-weighted neighborhood. Then, the gradient difference between the template image and inspected image was employed to determine the edge defect according to the grayscale difference threshold of weighted neighborhood. Finally, the difference artifacts were effectively eliminated by the two different image fusions and the real defects were retained. Experiments were conducted to compare the defect detection rate of printed image by using the traditional and proposed methods. Results demonstrate that for the most common dot defects the detection rate of the proposed methods is significantly higher than that of the traditional difference method due to the effective elimination of artifacts. The parallel acceleration based on compute unified device architecture (CUDA) enables the algorithm to speed up the defect detection of large print images by more than 60 times. The study provides significantly references for industrial inspection based on machine vision.
- Subjects
IMAGE processing; OPTOELECTRONIC devices; PHOTOELECTRIC devices; OPTICAL processors; IMAGE recognition (Computer vision); COMPUTER vision; OPTICAL devices
- Publication
Journal of Engineering Science & Technology Review, 2018, Vol 11, Issue 1, p180
- ISSN
1791-2377
- Publication type
Article
- DOI
10.25103/jestr.111.22