本研究應用影像處理技術及分類方法於液晶顯示器(liquid crystal display, LCD)偏光膜(polarizing film)的瑕疵檢測與分類。因偏光膜產業的競爭越來越激烈，然而目前產業界於偏光膜的自動化瑕疵檢測方面仍缺乏具有高分類準確率的檢測與分類系統。為取代容易誤判的人工分類過程，讓製造廠商可以迅速地得知製程中需要改善或維修的階段，因此本研究提出一套具有高分類準確率的檢測與分類系統，針對偏光膜之中常見的壓點(dents)、異物(particles)、亮點(bright spots)、刮傷(scratches)等四種瑕疵進行辨識與分類。
首先本研究利用中值濾波器(median filter)去除偏光膜瑕疵影像中的脈衝雜訊(impulse noise)，而針對背景中尚存的不規則雜訊則以改良型非等向擴散(anisotropic diffusion)予以平滑並同時銳化瑕疵區域的邊緣細節，接著以傅立葉轉換(Fourier transform)將瑕疵影像轉換至頻率域，配合巴特沃斯高通濾波器(butterworth high pass filter)進一步銳化瑕疵區域的邊緣細節，再利用傅立葉反轉換轉回至空間域完成影像強化之過程。而影像分割的部分是利用肯尼邊緣偵測器(Canny edge detector)先找出瑕疵區域的邊緣，再利用兩階段的形態學(morphology)處理以取得完整的瑕疵區域。對於四種常見瑕疵之分類，本研究是以瑕疵影像中所擷取到之灰階共生矩陣(gray level co-occurrence matrix, GLCM)中的對比度(contrast)、灰階共生矩陣中的均勻性(energy)、最大灰階值(maximum gray level)及偏心度(eccentricity)等四種特徵值作為分類器之輸入。最後利用徑向基函數類神經網路(radial basis function neural network, RBFNN)及倒傳遞類神經網路(back-propagation neural network, BPNN)作為分類器，使用96個瑕疵影像作為訓練樣本，再由184個瑕疵影像作為測試樣本以驗證分類效果，其中徑向基函數類神經網路的總體辨識率達到98.9%，證實可應用於產業界，藉此提升產品良率並降低成本。

In this study, image-processing techniques and classification methods were applied to propose an automated defect inspection and classification system of liquid crystal display polarizing films. Nowadays, competitions between polarizing films manufacturers are fiercer than before. Furthermore, the current automated defect inspection and classification systems used in companies manufacturing polarizing films still lack high classification accuracy. To replace the manual classifying process that easily misjudges and allow manufacturers to quickly know parts of the manufacturing process to fix, we proposed an automated defect inspection and classification system with very high classification accuracy. We focused on classifying four common defects: dents, particles, bright spots, and scratches.
During the image enhancement process, we applied median filter to eliminate background impulse noises. Then, improved anisotropic diffusion was used to soothe the irregular background noises and to sharpen the edge and detail of defect regions. To make the edge and detail sharper and background smoother, we then applied Fourier transform with butterworth high pass filter executed in frequency domain. To achieve precise image segmentation of defect regions, we used the Canny edge detector to find optimal edges of defect regions. Then, the two-stage morphology process was used to correctly determine the entire defect regions. According to the effectiveness analysis of different features, four types of input features for classifier, which were contrast of gray level co-occurrence matrix (GLCM), energy of gray level co-occurrence matrix, maximum gray level, and eccentricity, were extracted. With 96 defect images as training samples and 184 defect images as test samples, we finally used the radial basis function network (RBFNN) and back-propagation neural network (BPNN) as classifiers to validate the classification accuracy of our proposed system. The result shows that the classification accuracy by using the radial basis function network is 98.9%. Thus, our proposed system can be used by manufacturing companies for higher yield rate and lower cost.

[1] 葉錦清，「偏光膜產業發展趨勢分析」，工研院IEK產業情報網，四月，2011年。
[2] 葉仰哲，「LCD偏光板市場與廠商動態回顧」，工研院IEK產業情報網，十月，2013年。
[3] Y. G. Yoon, S. L. Lee, C. W. Chung, and S. H. Kim, “An Effective Defect Inspection System for Polarized Film Images Using Image Segmentation and Template Matching Techniques”, Computers & Industrial Engineering, Vol. 55, No. 3, pp. 567-583, 2008.
[4] C. H. Noh, S. L. Lee, D. H. Kim, C. W. Chung, and S. H. Kim, “An Effective and Efficient Defect Inspection System for TFT-LCD Polarized Films Using Adaptive Thresholds and Shape-Based Image Analyses”, International Journal of Production Research, Vol. 48, No. 17, pp. 5115-5135, 2010.
[5] 周宥澄，「智慧型線上即時偏光板瑕疵檢測系統之開發與研製」，國立台灣科技大學高分子工程研究所碩士論文，2007年。
[6] P. Perona, and J. Malik, “Scale-Space and Edge Detection Using Anisotropic Diffusion”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 12, No. 7, pp. 629-639, 1990.
[7] S. M. Chao, and D. M. Tsai, “Anisotropic Diffusion with Generalized Diffusion Coefficient Function for Defect Detection in Low-Contrast Surface Images”, Pattern Recognition, Vol. 44, No. 5, pp. 1917-1931, 2010.
[8] S. M. Chao, and D. M. Tsai, “An Improved Anisotropic Diffusion Model for Detail- and Edge-Preserving Smoothing”, Pattern Recognition, Vol. 31, No. 13, pp. 2012-2023, 2010.
[9] 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.
[10] S. Chen, B. Lin, X. Han, and X. Liang, “Automated Inspection of Engineering Ceramic Grinding Surface Damage Based on Image Recognition”, The International Journal of Advanced Manufacturing Technology, Vol. 66, No. 1, pp. 431-443, 2013.
[11] H. D. Lin, “Computer-Aided Visual Inspection of Surface Defects in Ceramic Capacitor Chips”, Journal of Materials Processing Technology, Vol. 189, No. 3, pp. 19-25, 2013.
[12] C. J. Lu, and D. M. Tsai, “Defect Inspection of Patterned TFT-LCD Panels Using A Fast Sub-Image Based SVD”, International Journal of Production Research, Vol. 42, No. 20, pp. 4331-4351, 2004.
[13] K. Singh, and R. Kapoor, “Image Enhancement Using Exposure Based Sub Image Histogram Equalization”, Pattern Recognition Letters, Vol. 36, pp. 10-14, 2014.
[14] H. O. Chen, N. S. P. Kong, and H. Ibrahim, “Bi-Histogram Equalization with A Plateau Limit for Digital Image Enhancement”, IEEE Transactions on Consumer Electronics, Vol. 55, No. 4, pp. 2072-2080, 2009.
[15] H. Hedberg, P. Dokladal, and V. Owall, “Binary Morphology with Spatially Variant Structuring Elements: Algorithm and Architecture”, IEEE Transactions on Processing, Vol. 18, No. 3, pp. 562-572, 2009.
[16] M. H. Shi, S. S. Jiang, H. R. Wang, and B. Xu, “Multiclass Defect Detection and Classification in Weld Radiographic Images Using Geometric and Texture Features”, Expert Systems with Applications, Vol. 37, No. 12, pp. 7606-7614, 2010.
[17] C. Y. Chang, C. H. Chang, C. H. Li, and M. D. Jeng, “Learning Vector Quantization Neural Networks for LED Wafer Defect Inspection”, International Journal of Innovative Computing, Vol. 4, No. 10, pp. 2565-2579, 2007.
[18] C. Y. Chang, C. H. Chang, C. H. Li, and M. Jeng, “Radial Basis Function Neural Networks for LED Wafer Defect Inspection”, Second International Conference on Innovative Computing, Information and Control, pp. 230-234, 2007.
[19] C. Y. Chang, C. H. Li, S. Y. Lin, and M. Jeng, “Application of Two Hopfield Neural Networks for Automatic Four-Element LED Inspection”, IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, Vol. 39. No. 3, pp. 325-365, 2009.
[20] 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.
[21] E. H. Land, “Display Device Employing Polarized Light”, United State Patent Office, No. 209061, 1938.
[22] 賴耀輝，「運用六標準差管理於LCD偏光膜維修之製程改善」，私立南台科技大學工業管理研究所碩士論文，2010年。
[23] 明基材料股份有限公司網站，偏光片研發技術，2014年，http://www.benqmaterials.com/chinese/02_rd/rd_detail.aspx?pid=2 。
[24] 李雅惠，「縮短偏光板製造業新產品開發送樣時程之研究-以X公司為例」，國立中央大學工業管理研究所碩士論文，2009年。
[25] 洪綺英，「二色性染料偏光膜製作」，國立台北科技大學有機高分子研究所碩士論文，2009年。
[26] 林玉堂，「偏光板產業安全文化評估」，國立高雄第一科技大學環境與安全衛生工程所碩士論文，2011年。
[27] J. Ma, X. Ye, and B. Jin, “Structure and Application of Polarizer Film for Thin-Film-Transistor Liquid Crystal Displays”, Displays, Vol. 32, No. 2, pp. 49-57, 2011.
[28] R. G. Gonzalez, and R. E. Woods, “Digital Image Processing 3rd Edition”, Prentice Hall, Taiwan, Taipei, 2008.
[29] J. W. Jukey, “Nonlinear (Nonsuperposable) Method for Smoothing Data”, 1974 Electronics and Aerospace Systems Conference, pp. 673, 1974.
[30] W. K. Pratti, “Median Filtering”, Meta-heuristic Intelligence Based Image Processing, Vol. 31, No. 13, pp. 2012-2023, 1975.
[31] 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.
[32] J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, No. 6, pp. 679-698, 1986.
[33] R. Medina-Carnicer, R. Munoz-Salinas, E. Yeguas-Bolivar, and L. Diaz-Mas, “A Novel Method to Look for The Hysteresis Thresholds for The Canny Edge Detector”, Pattern Recognition, Vol. 44, No. 6, pp. 1201-1211, 2011.
[34] J. Serra, “Image Analysis and Mathematical Morphology”, Academic Press, New York, 1982.
[35] 鍾國亮 編著，「影像處理與電腦視覺第五版」，東華書局有限公司，台灣台北，2014年。
[36] R. M. Haralick, “Statistical and Structural Approaches to Texture”, Proceedings of the IEEE, Vol. 67, No. 5, pp. 786-804, 1979.
[37] 蘇木春，張孝德 編著，「機器學習：類神經網路、模糊系統以及基因演算法則」，全華圖書，台灣台北，2007年。
[38] 王進德、蕭大全 編著，「類神經網路模式應用與實作」，全華圖書，台灣台北，2005年。
[39] 羅華強 編著，「類神經網路：MATLAB的應用」，高立圖書，台灣台北，2005年。
[40] 張斐章 編著，「類神經網路導論：原理與應用」，滄海圖書，台灣台北，2010年。
[41] A. McAndrew 編著，「數位影像處理」，高立圖書，台灣台北，2010年。