缺陷樣型是提供產品發生異常原因的關鍵線索，由分析缺陷樣型可以找到缺陷發生的原因。目前檢測方式已逐漸由人工目視轉換為自動化檢驗的方式，然而對於缺陷樣型的判斷仍以人工目視的方式來決定，但由於人工方式進行會因為經驗和分類習慣的不同造成判斷上的不一致，因此無法及時排除異常的情形。本研究發展一個自動缺陷分類系統，其中利用缺陷偵測的手法偵測自動化光學檢測機台檢測出來的缺陷影像，再萃取10個相關的缺陷特徵值，以分類與迴歸樹建立分類規則以增進分類的一致性及減少相關人力資源的投入，並且搭配特徵值相似度比對的結果，區分其他類型的缺陷，以改善分類樹與迴歸樹無法分類其他類的缺點。
為驗證效度，本研究與新竹科學園區一家半導體廠商為例進行實證，蒐集大宗缺陷樣型影像進行分析及驗證。實證結果顯示本研究所提出之方法能夠將缺陷樣型有效地分類，整體正確率達94%。

Yield improvement is an important issue in semiconductor manufacture industry. In the color filter process, it’s critical to identify the defect result through the defect map. However, automated optical inspection (AOI) equipment cannot classify the defect types, and the defect types are determined by human now. Thus, it would cause the classified result inconsistent. In this study, we classify the defect into eight categories, and used the canny edge detection to capture the defect, and find the features. Finally, we also used the classification and regression tree (CART) to develop the automatic defect classification.

[1] Breiman, L., J.H. Freidman, R.A. Olshen, C.J. Stone. 1984. Classification and Regression Trees, Wadsworth.
[2] Brunelli, R., T. Poggio. 1993. Face recognition: features versus templates. IEEE Transaction on Pattern Analysis and Machine Intelligence 15(10) 1042-1052.
[3] Canny, J. 1986. A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence 8(1) 679-698.
[4] Chou, P.B, A.R. Rao, M.C. Sturzenbecker, F.Y. Wu, V.H. Brecher. 1997. Automatic defect classification for semiconductor manufacturing. Machine Vision and Application 9 201-214.
[5] Goshtasby, A., S. H. Game, J. F. Bartholic. 1984. A two-stage cross correlation approach to template matching. IEEE Transaction on Pattern Analysis and Machine Intelligence 6(3) 374-378.
[6] Haralick, R.M., K. Shanmugam and I. Dinstein. 1973. Textural features for image classification. IEEE Trans vol. SMC-3 610-621 (1973).
[7] Kapur, J.N., P.K. Sahoo, A.K .C Wong. 1985. A New Method for Gray-Level Picture Thresholding Using the Entropy of the Histogram. Computer vision, Graphics and image processing 29 273-285.
[8] Lemon, S.C., Roy, J., Clark, M.A., Friedmann, P. D., & Rakowski, W. (2003). Classification and regression tree analysis in public health: Methodological review and comparison with logistic regression. Annals of Behavioral Medicine, 26(3), 172-181.
[9] Li, W.C., D.M. Tsai. 2011. Defect Inspection in Low-Contrast LCD Images Using Hough Transform-Based Nonstationary Line Detection. IEEE Transactions on Industrial Informatics 7(1) 136-147.
[10] Otsu, N. 1979. A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man and Cybernetics 9(1) 62-66.
[11] Shewhart, W.A. 1931. Economic Control of Quality of Manufactured Product. Van Nostrand, New York.
[12] Tseng, D.C., I.L. Chung, P.L. Tsai, C.M. Chou. 2011. Defect Classification for LCD Color Filters Using Neural-Network Decision Tree Classifier. International Journal of Innovative Computing, Information and Control 7(7A) 3695-3707.
[13] Tseng, D.C., c.y. Hung. 2005. Automatic defect inspection of patterned thin film transistor-liquid crystal display (TFT-LCD) panels using one-dimensional Fourier reconstruction and wavelet decomposition. International Journal of Production Research 43(21) 4589-4607.
[14] Wong, w.k., C.W.M. Yuen, D.D Fan, L.K Chan, E.H.K. Fung. 2009. Stitching defect detection and classification using wavelet transform and BP neural network. Expert Systems with Application 227(2) 3845-3856 (part2).
[15] Veltkamp, R.C. 2001. Shape Matching: Similarity Measures and Algorithms, invited talk. Proc. Int.l Conf. on Shape Modelling and Applications 188-197.