http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
조국한(Guk-Han Jo),허영욱(Young-wook Heo),송영준(Young-Joon Song) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
본 눈문에서는 결함 검출을 위해 딥러닝 학습데이터로 사용되는 패널이미지를 추출하는 방법에 대하여 설명한다. 패널의 이미지는 카메라와 Frame Graber를 이용하여 추출하며, 패널 포지셔닝, 휘도 설정과 같은 다양한 작업을 통하여 패널 이미지를 추출한다. 본 논문에서는 추출된 이미지 데이터들이 머신러닝 모델의 학습데이터로 사용될 수 있는지를 확인하기 위해 CNN 모델을 구성하고 결함 검출을 수행하였다. In this paper, a method of extracting a panel image used as deep learning learning data for defect detection is described. The image of the panel is extracted using the camera and Frame Grabber, and the panel image is extracted through various tasks such as panel positioning and luminance setting. In this paper, in order to confirm whether the extracted image data can be used as training data for the machine learning model, a CNN model was constructed and defect detection was performed.
LCD-Panel에서 MURA 검출을 위한 CNN 활용
조국한(Guk-Han Jo),조현우(Hyun-Woo Cho),허영욱(Youngwook Heo),송영준(Young-Joon Song) 대한전자공학회 2020 대한전자공학회 학술대회 Vol.2020 No.8
There are direct visual and machine vision methods to detect MURA in LCD panels. The direct visual method shows inconsistent results depending on the status of testers and poor MURA detection performance. The machine vision detection method is not good at detection of various forms of MURA. To solve these problems, this paper uses a representative model of machine learning, that is CNN(Convolutional Neural Network), to detect MURA. The CNN model used in this paper consists of five convolution layers and two fully connected layers. As a result of the experiment using the proposed method shows the 96% accuracy of MURA detection and takes 0.145 seconds for one panel test.