因目前3D顯示領域逐漸盛行，而其中光學薄膜部分因具有微結構，檢測上較一般薄膜困難，故本研究擬開發一套3D立體光學膜自動瑕疵檢測系統，欲檢測目前產線上所產生之瑕疵(直刻痕、橫刻痕、異物、壓點及紋理缺陷)。
本研究應用正交投影技術找尋光學薄膜上之微結構週期將影像自動定位，再使用相關係數法找出微結構之重覆週期當作標準模板，藉由相關係數法克服模板比對時產生之位移及旋轉問題，以影像相減法(Image Subtraction)將標準樣板與擷取影像相減並結合管制界限法(Control Limits Law)與物件標記法(Connected Component Labeling)分割出3D立體光學薄膜瑕疵位置，並且運用自組織特徵映射網路(Self-Organizing map, SOM)結合學習向量量化網路(Learning Vector Quantization, LVQ)輸入瑕疵特徵(面積、周長、形狀飽滿度、灰階值與長寬比)作為瑕疵分類系統；最後結合電腦人機控制介面、輸送平台、影像擷取設備、光源設備、馬達電控系統及影像處理之理論，以達到3D立體光學膜軟硬體整合之自動化瑕疵檢測目的。本研究其整體辨識率達94.4%，並可藉由檢測結果，回饋製程上之實際改善，亦經實務驗證於產線，確有助於提升檢測之效能，提高顯示器產業相關領域之市場競爭力。
關鍵字：相關係數、影像相減、管制界限法、自組織特徵映射網路、學習向量量化網路。

Three-dimensional (3D) display has become increasingly prevalent and the detection of 3D display optical film is more difficult than general thin films due to the microstructures. This study attempted to develop a 3D optical film automatic defect detection system for the detection of the defects on the current production line (straight notch, transverse notch, foreign object, pressing point and texture defect).
This study applied the orthogonal projection technology to identify the microstructure cycles on the optical thin film for the automatic location of images. Then, the correlation coefficient approach was used to determine the repeated cycles of the microstructures as the standard templates, and overcome the displacement and rotation problems generated in template matching using the correlation coefficient approach. The image reduction approach was employed to subtract the standard templates and the captured images to segment the 3D optical thin film defects by combining the control limit method and object labeling method. This study also applied the self-organizing feature mapping network, combined with learning quantitative network, to input defect features (area, perimeter, shape fullness, grey scale value and aspect ratio) as the defect classification system. Finally, the computer-human control interface, delivery platforms, image capture equipment, lighting equipment, motor electrical control system were combined with image processing theories to realize the purpose of automatic defect detection of 3D optical film by integrated software and hardware. The overall detection rate of the proposed system reached 94.4%. The detection results could be fed back to manufacturing process for practical improvements. As proved by application in the production line, the proposed system can help to improve the detection performance, and enhance the market competitiveness in the display-related fields.
Keyword: Correlation coefficient approach、Image reduction approach、Control limit method、Self-organizing feature mapping network、Learning quantitative network.

[1] C. Neubauer, “Segmentation of defects in textile fabric”, International Conference on Computer Vision and Applications, Hague, pp. 688-691(1992).
[2] A.L. Amet, A. Ertuzun, and A. Ercil, “Texture defect detection using subband domain co-occurrence matrices”, International Conference on Image Analysis and Interpretation, Tucson, pp. 205-210(1998).
[3] D. M. Tsai, S. T. Chuang, and Y. H. Tseng, “One-dimensional based automatic defect inspection of multiple patterned TFT-LCD panels using Fourier image reconstruction”, International Journal of Production Research, Vol. 45, No. 6, pp. 1297-1321(2007).
[4] D. M. Tsai, and C. Y. Hsieh, “Automated surface inspection for directional textures”, Image and Vision Computing, Vol. 18, No. 1, pp. 49-62(1999).
[5] A. Kumar, and G. Pang, “Identification of surface defects in textured materials using wavelet packets”, Industry Applications Conference, Chicago, Vol. 1, pp. 247-251(2001).
[6] X. Yang, G. Pang, and N. Yung, “Robust fabric defect detection and classification using multiple adaptive wavelets”, IEEE Proceedings Vision, Image and Signal Processing, Vol. 152, No. 6, pp. 715-723(2005).
[7] S. Yoshimura, and T. Kanade, “Fast template matching based on the normalized correlation by using multiresolution eigenimages”, International Conference on Intelligent Robots and Systems, Munich, Vol. 3, pp. 2086-2093(1994).
[8] M. Uenohara, and T. Kanade, “Use of Fourier and Karhunen-Loeve decomposition for fast pattern matching with a large set of templates”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 19, No. 8, pp. 891-898(1997).
[9] V. Trappey, H. Y. Wu, J.C. Trappey, and T. D. Fataneh, “Using patent data for technology forecasting: China RFID patent analysis”, Advanced Engineering Informatics, Vol. 25, No. 1, pp. 53-64(2011).
[10] D. M. Tsai, and Y. H. Tsai, “Defect detection in textured surfaces using color ring-projection correlation”, Machine Vision and Applications, Vol. 13, No. 4, pp. 194-200(2003).
[11] S. Kaneko, Y. Satoh, and S. Igarashi, “Using selective correlation coefficient for robust image registration”, Pattern Recognition, Vol. 36, No. 5, pp. 1165-1173(2003).
[12] J. C. Yoo, and T. H. Han, “Fast normalized cross correlation”, International Journal of Circuits, Systems and Signal Processing, Vol. 28, No. 6, pp. 819-843(2009).
[13] 林京佑， “改良式相關係數法於晶粒表面瑕疵之檢測”，元智大學，工業工程與管理研究所，碩士論文(2006)。
[14] 吳俊樟， “小波轉換於圖形比對的一種方法”，中央大學，數學研究所，碩士論文(2007)，。
[15] T. W. Ridler, and S. Calvard, “Picture thresholding using an iterative selection method”, IEEE Transactions on Systems, Vol. 8, No. 8, pp. 630-632(1978).
[16] N. Otsu, “A threshold selection method from gray level histogram”, IEEE Transactions on Systems, Vol. 9, No. 1, pp. 62-66(1979).
[17] L. Cao, Z. Shi, and E. K. W. Cheng, “Fast automatic multilevel thresholding method”, Institution of Engineering and Technology, Vol. 38, No. 16, pp. 868-870(2002).
[18] D. M. Tsai, and C. H. Yang, “A quantile–quantile plot based pattern matching for defect detection”, Pattern Recognition Letters, Vol. 26, No. 13, pp. 1986-1962(2005).
[19] M. Frucci, G. Ramella, and G. Baja, “Using resolution pyramids for watershed image segmentation”, Image and Vision Computing, Vol. 25, No. 6, pp. 1021-1031(2007).
[20] P. Hu, Y. Luo, and C. Li, “Chinese chess recognition based on projection histogram of polar coordinates image and FFT”, Chinese Conference on Pattern Recognition, Nanjing, pp. 1-5(2009).
[21] T. Cover, and P. Hart, “Nearest neighbor pattern classification”, IEEE Transactions on Information Theory, Vol. 13, No. 1, pp. 21-27(1967).
[22] G. Barnard, and T. Bays, “Studies in the history of probability and statistics”, Oxford Journal of Biometrika, Vol. 59, No. 2, pp. 239-251(1972).
[23] F. L. Chen, and S. F. Liu, “A neural-network approach to recognize defect spatial pattern in semiconductor fabrication”, IEEE Transactions on Semiconductor Manufacturing, Vol. 13, No. 3, pp. 366-373(2000).
[24] J. Vesanto, and E. Alhoniemi, “Clustering of the self-organizing map”, IEEE Transactions on Neural Networks, Vol. 11, No. 3, pp. 586-600(2000).
[25] G. Acciani, G. Brunetti, and G. Fornarelli, “Application of neural networks in optical inspection and classification of solder joints in surface mount technology”, IEEE Transaction on Industrial Informatics, Vol. 2, No. 3, pp. 200-209(2006).
[26] 莊東漢，平面顯示器概論，高立圖書有限公司(2008)。
[27] O. Schreer, P. Kauff, and T. Sikora,“3D Videocommunication Algorithms concepts and real time systems in human centred communication”, John Wiley & Sons Ltd (2005).
[28] 陳威仰， “適應性相減法於週期性紋路之表面瑕疵檢測”，元智大學，工業工程與管理學系，碩士論文(2008)。
[29] 鐘國亮，影像處理與電腦視覺，東華書局有限公司(2008)。
[30] Gonzalez Woods, “Digital Image Processing 3/e”, Prentice Hall(2009).
[31] 蘇木春、張效德，機器學習：類神經網路、模糊系統以及基因演算法則，全華圖書股份有限公司(2007)。
[32] T. Kohonen, “Self-organized formation of topologically correct feature maps”, Biological Cybernetics, Vol. 43, No. 1, pp. 59–69(1982).
[33] 繆紹剛，數位影像處理-運用MATLAB，東華書局股份有限公司(2005)。
[34] 連國珍，數位影像處理MATLAB，儒林圖書公司(2007)。
[35] 羅華強，類神經網路-MATLAB的應用，高立圖書有限公司(2005)。
[36] Carsten Steger, Markus Ulrich, and Christian Wiedemann, “Machine Vision Algorithms and Applications“, Photon-Tech Instruments Co. Ltd(2011)。