本研究主要針對薄膜電晶體液晶平面顯示器(thin film-liquid crystal display, TFT-LCD)模組瑕疵進行辨識與分類，檢測TFT-LCD模組瑕疵種類分為異物(foreign matters)、暗點(dark defect)、丸狀Mura(pelletized Mura)、線缺陷(line defect)等四種瑕疵。首先，對於TFT-LCD模組的週期性紋理，本研究首先利用傅立葉轉換(Fourier transform)找出正確週期並且利用低通濾波器(low-pass filters)來消除週期性紋理以利於瑕疵辨識，低通濾波後之影像雜訊再利用非等向擴散(anisotropic diffusion)來消除雜訊並且銳化了瑕疵的部分，而對於影像光源不均的問題，本研究利用迴歸分析(regression analysis)法預測影像的灰階分佈，利用原影像與預測的影像作影像相減，並且利用中值濾波消除雜訊及二值化(Otsu)分割出瑕疵區域，最後再以形態學(morphology)的方法保留瑕疵完整結構，針對擷取的瑕疵面積大小(size)、瑕疵長寬比(aspect ratio)、似圓性(circularity)及形狀飽和度(compactness)此四種特徵值作為分類器之輸入。實務驗證上，本研究所提出之瑕疵檢測方法，透過開發之軟體，可於3.02秒完成瑕疵辨識。
在瑕疵分類上，將擷取之四種特徵值利用C4.5決策樹(decision tree)及倒傳遞類神經(back-propagation neural, BPN)網路作為瑕疵之分類決策理論。實務驗證下，本研究利用55個瑕疵作為訓練樣本，再將130個測試樣本之特徵值作為兩者分類器之輸入，以C4.5決策樹作為分類器，其總體辨識率為93.8%；以倒傳遞分類器，其總體辨識率為98.5%，證實本研究提出的C4.5決策樹分類法能夠應用自動化光學檢測技術於TFT-LCD模組瑕疵分類，成功的提升生產良率以降低成本。

This study identified and classified the thin film-liquid crystal display (TFT-LCD) module defects. The defects of the TFT-LCD modules were divided into four types, namely, foreign matter, dark defect, pelletized Mura, and line defect. First, for the periodic texture of TFT-LCD module, this study used the Fourier transform to find the correct cycle, and used an low-pass filter to eliminate the periodic texture to facilitate the defect identification. The image noise after the low-pass filtering was eliminated by anisotropic diffusion, which also sharpened the defective parts. Regarding the problem of uneven light source of image, this study used the regression analysis to predict the gray level distribution of the images. The image subtraction was conducted by using the original image and the predicted image. The noise was eliminated by the median filter and the defect areas were partitioned using the Euclidean method. Finally, the complete structure of defects were preserved with the morphological method. The five captured eigenvalues, namely size, aspect ratio, circularity and compactness of the defects were used as the classifier inputs. Through validation, the proposed defect detection method could complete the detection in 3.02 sec through the developed software.
In the defect classification, the five kinds of eigenvalues captured were used as the defect classification decision theory using the C4.5 decision tree and back-propagation neural network (BPNN). In experimental validation, 55 defects were used as the training samples, and eigenvalues of 130 test samples were input into the two classifiers. When using C4.5 decision tree as the classifier, the overall identification rate was 93.8%; when using the BPNN classifier, the overall identification rate was 98.5%. The results proved that the C4.5 decision tree classification method proposed in this study can successfully apply the automated optical inspection technology on the TFT-LCD module defect classification, thus improving the yield rate and reducing the cost.

[1] 王木俊 編著，「薄膜電晶體液晶顯示器－原理與實務」，新文京開發出版股份有限公司，台灣台北，2008年。
[2] 林宸生 編著，「光電科技與生活－光電科技導論」，五南圖書出版公司，台灣台北，2008年。
[3] 石士欣，「以Haar小波轉換與空間濾波於TFT-LCD面板之瑕疵檢測」，國立台灣科技大學電機工程研究所碩士論文，2009年。
[4] 陳思宏，「彩色濾光片Mura瑕疵之瑕疵檢測」，義守大學工業工程管理學系研究所碩士論文，2006年。
[5] 林冠良，「應用灰階共生矩陣於彩色濾光片瑕疵檢測之研究」，義守大學工業工程管理學系研究所碩士論文，2009年。
[6] R. M. Haralick, K. Shanmugam, and I. H. Dinstein,“Textural features for image classification,”IEEE Transactions on Systems, Man and Cybernetics, Vol. 3, No. 6, pp. 610-621, 1973.
[7] S. K. S. Fan, and Y. C. Chuang,“Automatic detection of Mura defect in TFT-LCD based on regression diagnostics,”Pattern Recognition Letters, Vol. 33, No. 15, pp. 2397-2404, 2010.
[8] D. M. Tsai, and W. C. Li,“Defect inspection in low-contrast LCD Images using Hough transform-based Non-stationary line detection,”IEEE Transactions on Industrial Information, Vol. 7, No. 1, pp. 136-147, 2011.
[9] S. L. Chen, and J. H. Chang,“TFT-LCD Mura defects automatic inspection system using linear regression diagnostic model,”Journal of Engineering Manufacture, Vol. 222, No. 11, pp. 1489-1501, 2008.
[10] N. K. Park, and S. I. Yoo,“Evaluation of TFT-LCD defects based on human visual perception,”Displays, Vol. 7, No. 1, pp. 1-16, 2009.
[11] S. M. Chao, and D. M. Tsai,“An anisotropic diffusion-based defect detection for low-contrast glass substrates,”Image and Vision Computing, Vol. 26, No. 2, pp. 187–200, 2008.
[12] S. M. Chao, and D. M. Tsai,“Anisotropic diffusion with genera-lized diffusion coefficient function for defect detection in low-contrast surface images,”Pattern Recognition, Vol. 43, No. 5, pp. 1917-1931, 2010.
[13] P. Perona, J. Malik,“Scale-space and edge detection using aniso-tropic diffusion,”IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 12, No. 7, pp. 629–639, 1990.
[14] S. M. Chao, and D. M. Tsai,“An improved anisotropic diffusion model for detail- and edge-preserving smoothing,”Pattern Recognition Letters, Vol. 31, No. 13, pp. 2012–2023, 2010.
[15] U. K. Nafis,“Histogram statistics based variance controlled adaptive threshold in anisotropic diffusion for low contrast image enhancement,”Signal Processing, Vol. 93, No. 6, pp. 1684–1693, 2013.
[16] L. C. Chen, and C. C. Kuo,“Automatic TFT-LCD Mura defect inspection using discrete cosine transform-based background filteringand ‘just noticeable difference quantification strategies,”Measurement Science and Technology, Vol. 19, No. 1, 2008.
[17] L. C. Chen, C. H. Chen, and X. L. Nguyen,“An effective image segmentation method for noisy low-contrast unbalanced background in Mura defects using balanced discrete-cosine-transfer (BDCT),” Precision Engineering, Vol. 37, No. 2, pp.336-344, 2013.
[18] 王敬欽，「應用自動化光學檢測技術於彩色濾光片微觀瑕疵辨識系統之開發與研究」，國立台灣科技大學自動化及控制研究所碩士學位論文，2009年。
[19] C. F. Kuo, C. T. Hsu, C. H. Fang, S. M. Chao, and Y. D. Lin, “Automatic defect inspection system of color filters using Tagu-chi-based neural network,”International Journal of Production Research, Vol. 51, No. 5, pp. 1464-1476, 2013.
[20] H. D. Lin,“Automated defect inspection of light-emitting diode chips using neural network and statistical approaches,”Expert Systems with Applications, Vol. 36, No. 1, pp. 219-226, 2009.
[21] S. Ravikumar, K. I. Ramachandran and V. Sugumaran,“Machine learning approach for automated visual inspection of machine components,”Expert Systems with Applications, Vol. 38, No. 4, pp. 3260-3266, 2011.
[22] J. E. Harry,“Introduction to plasma technology : science, engineering and applications,”Wiley-VCH, USA, 2010.
[23] H. Z. Kafafi,“Organic electroluminescence,”Boca Raton, USA, 2005.
[24] 張頌榮，「TFT-LCD面板之點線瑕疵檢測自動化檢測系統」，國立成功大學製造工程研究所碩士論文，2005年。
[25] 丁琨洲，「應用於TFT-LCD Panel檢驗之測試樣本產生器雛形設計」，南台科技大學電子工程研究所碩士論文，2007年。
[26] K. Nakashima,“Hybrid inspection system for LCD color filter panels,”Proceedings 10th Anniversary IEEE Instrumentation and Measurement Technology Conference, Hamamatsu, pp. 689-692, 1994.
[27] SEMI D31-1102, Definition of measurement index (SEMU) for luminance MURA in FPD image quality inspection.
[28] 張瑞顯，「應用線性迴歸診斷法於液晶顯示器Mura缺陷自動化檢測設計與實現」，國立成功大學製造工程研究所碩士論文，2005年。
[29] R. G. Gonzalez, and R. E. Woods,“Digital image processing 3rd edition prentice hall,”Taipei, Taiwan, 2008.
[30] B. Tamal,“Digital signal and image processing,”WILEY, USA, 2004.
[31] 吳宗正 編著，「迴歸分析」，三民書局，台灣台北，1993年。
[32] 葉怡成 編著，「高等實驗計畫法」，五南文化事業，台灣台北，2009年。
[33] G. Sawitzki,“Computational statistics : an introduction to R,”Chapman and Hall, Heidelberg, Germany, 2009.
[34] J. N. Kapur,“New method for gray-level picture thresholding using the entropy of the histogram,”Computer Vision, Graphics, and Image Processing, Vol. 29, No. 3, pp.273-285, 1985.
[35] Z. Li, C. Liu, G. Liu, Y. Cheng, X. Yang, and C. Zhao,“A novel statistical thresholding method,”Int. J. Electron. Commun., Vol. 64, No. 12, pp. 1137-1147, 2010.
[36] N. Otsu,“A threshold selection method from gray-level histograms,” IEEE Transactions on Systems, Man and Cybernetics, Vol. 9, No. 1, pp. 62-66, 1979.
[37] 鍾國亮 編著，「影像處理與電腦視覺－第五版」，東華書局，台灣台北，2012年。
[38] J. R. Quinlan,“Induction of decision trees,”Machine Learning, Vol. 1, No. 1, pp. 81-106, 1985.
[39] J. R. Quinlan,“C4.5 programs for machine leaning,”Morgan Kaufmann Publishers, San Mateo, CA, 1993.
[40] 陳士杰，「決策樹學習」，國立聯合大學資訊管理學系，2005年。
[41] 梁梓遠，「應用自動化光學檢測技術於發光二極體微觀瑕疵辨識系統之開發」，國立台灣科技大學自動化及控制研究所碩士學位論文，2011年。
[42] 羅華強 編著，「類神經網路：MATLAB的應用」，高立圖書，台灣台北，2005年。
[43] 張斐章 編著，「類神經網路導論：原理與應用」，滄海書局，台灣台北，2010年。
[44] A. McAndrew 編著，「數位影像處理」，高立圖書，台灣台北，2010年。