Journal
IEEE ACCESS
Volume 7, Issue -, Pages 105721-105733Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/ACCESS.2019.2931194
Keywords
Machine vision; industrial defect measurement; polarizer; transparent microdefect
Categories
Funding
- National Science Foundation of China [61571306]
- Shenzhen Fundamental Research fund [JCYJ20160331185006518, JCYJ20180306174120445]
- Startup Research Fund of Shenzhen University [2019041]
Ask authors/readers for more resources
The main task of vision-based industrial defect inspection is to implement efficient non-contact visual quality control, i.e., to detect if there is a defect and to achieve an accurate 3D shape measurement of such a defect, and this kind of vision defect inspection system has been widely applied in various industrial application. However, it is still not the case in the inspection of transparent microdefect on the polarizer (which is the most important part of an LCD screen). Optical measurement devices (such as confocal microscopy) are often utilized to fulfil this task. To solve problems lied in the current confocal microscopy inspection system, such as expensive and non-real-time processing, this research aims to develop a novel vision-based 3D shape measurement system for polarizer transparent microdefect characterization. The innovation of this system, which has been verified by our optical model simulation, is that the 3D sizes of microdefect have a monotonically relation to the grayscale of the microdefect image. Hence, a microdefect imaging system, which could acquire defect image accurately, is first well designed and implemented. Then, a support vector regression (SVR) algorithm is derived by the trained data, i.e., 100 acquired defect images and its corresponding 3D shape value by confocal microscopy. Characterized 3D measurement of microdefect is thereby obtained by this SVR algorithm. 30 polarizer microdefect samples have been imaged and measured by our proposed system, and several important performance indicators, including processing speed, accuracy and system reproducibility, have been elaborately tested. The experimental results show that the proposed system could achieve a high-accuracy measurement but in a much faster and more efficient way than the confocal microscopy. Besides, this developed imaging system has been evaluated in real applications, and over 300 samples have been detected, which also validate the effectiveness of the proposed system.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available