In the manufacture of flat display panels, salt-and-pepper defects are caused by a malfunction in the chemical process. The defects are characterized by the dispersion of many black and white pixels in the display panels; these pixels are difficult to detect with conventional automatic fault detection methods that specialize in recognizing certain shapes, such as line or mura defects (stains). This study proposes a simple but high-performance salt-and-pepper defect detection method. First, the background image of the original image is generated using the mean filter in the spatial domain to create a noise image, which is the subtraction of the two images. A binary image is then obtained from the noise image to count the defective pixels, and a statistical control chart that monitors the number of defective pixels identifies the panel defects. Two experiments were conducted with images collected from an organic light-emitting diode inspection process, and the proposed method showed excellent performance with respect to classification accuracy and processing time.
Bibliographical noteFunding Information:
Acknowledgements This work was supported in part by Samsung Display Co., Ltd., South Korea, by the Technology Innovation Program (10045913, Development of Big Data-Based Analysis and Control Platform for Semiconductor Manufacturing Plants) funded by the Ministry of Trade, Industry & Energy (MOTIE, South Korea), by the National Research Foundation (NRF) of Korea Grant funded by the Korean government (MSIP) (NRF-2016R1A2B4008337), and by the Global PhD Fellowship Program through the NRF of Korea funded by the Ministry of Education (NRF-2015H1A2A1031081).
© 2017, Springer Science+Business Media New York.
All Science Journal Classification (ASJC) codes
- Industrial and Manufacturing Engineering
- Artificial Intelligence