A Mura has been one of the most important defects in the display since Samsung Display Co., Ltd. was founded. Due to the variation in fabrication (FAB) process, the Mura is occurred. The only way of detecting Mura until now is judging by the human's recognizing after the panel is created. In the respect of manufacturing, late detection of it is a huge problem. In this research, we developed the system which predicts the Mura index in early step. This system extracts the Thin Film Transistor (TFT) characteristics, Resistance and Capacitance (RC) parameters and Optic information from images through image processing; to enhance the quality of images, upscaling method is introduced. High speed is essential for this monitoring system to be applied to manufacturing process; however it takes long time when using simulation. Therefore, we introduced machine learning (ML) for TFT characteristics and Optics as well as gradual meshing for RC extraction. Attributed to the system, the lead time for Mura detection is decreased from 10 days to 4days in OLED device (in LCD device, from 15 days to 2 days). And accuracy for the predictions is achieved over 0.80.

No abstract