本研究『應用影像處理技術及最佳化多類支持向量機於發光二極體晶粒表面色差與微瑕疵檢測系統之開發與研究』主要針對晶粒(Chip)製程的檢測階段進行自動化系統之最佳化設計，包含晶粒色差(Color Aberration)影像處理、晶粒定位、瑕疵特徵擷取，以及瑕疵分類；所檢測之晶粒瑕疵依區域可分為外型瑕疵(晶粒表面崩損)、發光區瑕疵(發光區表面顏色不一致，發光區破損)與電極區瑕疵(電極刮傷，電極汙染，電極無針痕，Finger斷裂)三個部分。
本研究利用形狀向量相關係數法的模板匹配(Shape-Base Template Matching)進行色差影像晶粒定位，並利用計算出的形狀相關係數做為特徵，對表面結構毀損的晶粒加以區分。晶粒定位後利用高斯平滑最小門檻值分割法(Gaussian Smoothing Minimal Threshold)、Otsu門檻值分割法(Otsu Threshold Segmentation Method)、型態學與影像相減等技術進行瑕疵標記，並計算瑕疵特徵，包含二值化特徵與灰階值特徵。最後利用田口方法(Taguchi Methods)結合主成分分析(Principal Component Analysis, PCA)進行特徵值分析，縮減分類的資料量與維度，以分析後的瑕疵特徵做為決策樹支持向量機(Decision Tree Support Vector Machine, DTSVM)分類模組的訓練依據，建立最佳化多類支持向量機分類模組(Multi-Class Support Vector Machine Model)針對電極區與發光區進行微瑕疵細分，並與傳統二元化結構支持向量機、類神經網路等分類器比較。結果顯示，本研究之檢測系統整體辨識率在96%以上且500顆晶粒瑕疵分類速度為3秒，驗證本研究系統有效定位晶粒以及獲得可靠瑕疵資訊，成功克服自動化瑕疵檢測系統在色差影像上的不足，建構出即使受到干擾仍然具有強健性的檢測流程，而且對瑕疵分類具有快速、高精準度與穩定度等特點，能有效應用於大量生產之發光二極體精密檢測，取代人工目檢來節省人力成本。

This study entitled “The Development and Research on Using Image Processing Technology and Optimization of Multi-Class Support Vector Machines in Light-emitting Diode (LED) Chip Surface Color Aberration and Micro-defect Detection System” aimed to propose the optimization design of the automatic system at the detection stage of the chip manufacturing process, covering LED chip color aberration image processing, chip location, defective feature extraction. Based on the defect, the defects of the tested chip can be divided into shape defects (chip surface collapse and loss), the light emitting area defects (light emitting area surface color differences, the light emitting area damage) and the electrode area defects (electrode scratching, electrode contamination, lack of pin marks in the electrode, finger fracture) by area.
This study used the shape-base template matching for color aberration image chip location, and used the calculation result of the shape correlation coefficient as the feature to distinguish the chips of damaged surface structure. After the chip location, this study employed the Gaussian smoothing minimal threshold segmentation approach, Otsu threshold segmentation method, and morphology and image subtraction to mark the defects, and calculate the defective features, including the binarized features and the grayscale features. Finally, the Taguchi method and the principal component analysis method were applied to analyze the defective features, reduce the classification data amount and dimensions. With the analyzed defect characteristics as the training basis for decision tree support vector machine, this study established the optimization multi-class support vector machines classification module to further distinguish the micro defects in the electrode area and the light emitting area. Comparison was made with the traditional binary structure support vector machines, neural network classifiers. The results suggested that the overall recognition rate of the inspection system can be more than 96% and the classification of defects in 500 chips takes only 3 sec, verifying that the system can effective locate the chip and obtain reliable defect information. The design can successfully overcome the shortcoming of the automatic defect inspection system in terms of color aberration image to establish a robust inspection procedure even under interference. Meanwhile, the defect classification is characterized by fast speed, high accuracy and high reliability. Therefore, it can be effectively applied in the precision inspection of mass-produced LED chips to replace the existing human eye inspection to save labor costs.

[1] 郭浩中、賴芳儀、郭守義，LED原理與應用，五南圖書出版股份有限公司，2009。
[2] Chang, C. Y., Chang, Y. C., Li, C. H., Jeng, M. D., “Radial Basis Function Neural Networks for LED Wafer Defect Inspection”, Second International Conference on Innovative Computing, Information and Control, pp. 230-230, 2007.
[3] Wu, J. Y., “Apply Image Processing Technology to the Appearance Defect Inspection of High Brightness Light Emitting Diode”, National Taiwan University of Science and Technology, Master thesis, 2007.
[4] Chang, C. Y., Chang, C. H., Li, C. H., and Jeng, M. D., “Learning Vector Quantization Neural Networks for LED Wafer Defect Inspection”, International Journal of Innovative Computing, Information and Control, Vol. 4, No. 10, pp. 2565-2579, 2008.
[5] Liao, Y. Y., “Apply Machine Vision Technology to Defect Inspection of High Brightness Light Emitting Diode”, National Taiwan University of Science and Technology, Master thesis, 2008.
[6] Lin, H. D., “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.
[7] Li, C. H., Chang, C. Y., Jeng, M. D., and Jeng , Y. T., “A Case Study of Applying Regional Level-Set Formulation to Post-Sawing LED Wafer Inspection”, IEEE International Conference on Systems Man and Cybernetics, pp. 2745-2752, 2010.
[8] Perng, D. B., Liu, H. W., and Chang, C. C., “Automated SMD LED Inspection Using Machine Vision”, International Journal of Advanced Manufacturing Technology, pp. 1-13, 2011.
[9] Tsai, D. M., Chiang, I. Y., and Tsai, Y. H., “A Shift-Tolerant Dissimilarity Measure for Surface Defect Detection”, IEEE Transactions on Industrial Informatics, Vol. 8, No. 1, pp. 128-137, 2012.
[10] Tsai, D. M., and Lin, C. T., “Fast Normalize Cross Correlation for Defect Detection”, Pattern Recognition Letters, Vol. 24, pp. 2625-2631, 2003.
[11] Kaneko, S., Satoh, Y., and Igarashi, S., “Using Selective Correlation Coefficient for Robust Image Registration”, Pattern Recognition, Vol. 36, No. 5, pp. 1165-1173, 2003.
[12] Gavrila, D.M., “A Bayesian, Exemplar-Based Approach to Hierarchical Shape Matching”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, No. 8, pp. 1408-1421, 2007.
[13] Zhang, K., Jiangbo, L., Lafruit, G., Lauwereins, R., and Luc, V. G., “Robust Stereo Matching With Fast Normalized Cross-Correlation Over Shape-Adaptive Regions”, IEEE International Conference on Image Processing, pp. 2357 – 2360, 2009.
[14] Huang, C. J., Wu, C. F., and Wang, C. C., “Image Processing Techniques for Wafer Defect Cluster Identification”, IEEE Design & Test of Computers, Vol. 19, No. 2, pp. 44-48, 2002.
[15] Cao, L., Shi, Z. K., and Cheng, E. K. W., “Fast Automatic Multilevel Thresholding Method”, Electronics Letters, Vol. 38, No. 16, pp. 868-870, 2002.
[16] Ward, G., “Fast Robust Image Registration for Compositing High Dynamic Range Photographs from Handheld Exposure”, Journal of Graphics Tools, Vol. 8, No. 2, pp. 17-30, 2003.
[17] Lu, D., Mausel, P., Brondizio, E., and Moran, E., “Change Detection Techniques”, International Journal of Remote Sensing, Vol. 25, No. 12, pp. 2365-2401, 2004.
[18] Engin, A., Derya, A., “An Expert System Based on Fuzzy Entropy for Automatic Threshold Selection in Image Processing”, Expert Systems with Applications, Vol. 36, pp. 3077-3085, 2009.
[19] Syrcos, G. P., “Die Casting Process Optimization Using Taguchi Methods”, Journal of Materials Processing Technology, Vol. 135, pp. 68-74, 2003.
[20] Nandi, G., Datta, S., Bandyopadhyay, A., and Pal P. K., “Application of PCA-Based Hybrid Taguchi Method for Correlated Multicriteria Optimization of Submerged Arc Weld: A Case Study”, International Journal of Advanced Manufacturing Technology, Vol. 45, pp. 276-286, 2009.
[21] Rajkiran, G., and Vijayan, K. A., “An Improved Face Recognition Technique Based on Modular PCA Approach”, Pattern Recognition Letters, Vol. 25, pp. 429-436, 2004.
[22] Pinto, R. C., Bosc, V., Nocairi H., Barros, A. S., and Rutledge, D. N., “Using ANOVA-PCA for Discriminant Analysis: Application to The Study of Mid-infrared Spectra of Carraghenan Gels as A Function of Concentration and Temperature”, Analytica Chimica Acta, Vol. 629, pp. 47-55, 2008.
[23] Kim, H. C., Pang, S., Je, M. H., Kim, D., and Bang, S. Y., “Pattern Classification Using Support Vector Machine Ensemble”, Proceeding 16th IEEE International Conference on Pattern Recognition, Vol. 2, pp. 160-163, 2002.
[24] Yang, X., Pang, G., and Yung, N., “Discriminative Training Approaches to Fabric Defect Classification Based on Wavelet Transform”, Pattern Recognition, Vol. 37, No. 5, pp. 889-899, 2004.
[25] Mulay, S. A., Devale, P. R., and Garje, G. V., “Decision Tree Based Support Vector Machine for Intrusion Detection”, IEEE International Conference on Networking and Information Technology, pp. 59 -63, 2010.
[26] Chang, F., Guo, C. Y., Lin, X. R., and Lu, C. J., “Tree Decomposition for Large-Scale SVM Problems”, Journal of Machine Learning Research, Vol.11, pp. 2855-2892, 2010.
[27] Azadeh, O., and Nasser, S., “An Efficient Hybrid Algorithm for Multi-Class Support Vector Machines”, International Conference on Computer Science and Information Technology, pp. 24-28, 2011.
[28] Carsten, S., Markus, U., and Christian, W., Machine Vision Algorithms and Applications, Wiley-VCH, 2011.
[29] Frank, E., Hall, M. A., Holmes, G., Kirkby, R. B., Pfahringer, B., and Witten, I. H., Data Mining and Knowledge Discovery Handbook, Springer, pp. 1305-1314, 2005.
[30] Chang, C. C., and Lin, C. J., “LIBSVM: A Library for Support Vector Machines”, http://www.csie.ntu.edu.tw/~cjlin/libsvm/, 2012.
[31] Nakamura, S., Senoh, M., and Mukai, T., “High-Power InGaN/GaN Double-Heterostructure Violet Light-Emitting Diodes”, Applied Physics Letters, Vol. 62, No. 19, pp. 2390-2392, 1993.
[32] Gonzalez, W., Digital Image Processing 3/e, Prentice Hall, 2008.
[33] Canny, J., “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, pp. 679-698, 1986.
[34] Tanimoto, S. L., “Template Matching in Pyramids”, Computer Graphics and Image Processing, Vol. 16, pp. 356-369, 1981.
[35] Glazer, F., Reynolds, G., and Anandan, P., “Scene Matching by Hierarchical Correlation”, In Computer Vision and Pattern Recogntion, pp. 432-441, 1983.
[36] Anisimov, V. A., and Gorsky, N. D., “Fast Hierarchical Matching of An Arbitrarily Oriented Template”, Pattern Recognition Letters, Vol. 14, pp. 95-101, 1993.
[37] Forstner, W., “A Framework for Low Level Feature Extraction”, Third European Conference on Computer Vision, Lecture Notes In Computer Science, Berlin, Vol. 801, pp. 383-394, 1994.
[38] Schmid, C., Mohr, R., and Bauckhage, C., “Evaluation of Interest Point Detectors”, International Journal of Computer Vision, Vol. 37, pp. 151-172, 2000.
[39] Otsu, N., “A Threshold Selection Method from Gray Level Histogram”, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, pp. 62-66, 1979.
[40] Bouaynaya, N., “Theoretical Foundations of Spatially-Variant Mathematical Morphology Part II: Gray-level Images”, IEEE Transactions on Pattern Analysis and Maxhine Intelligence, Vol. 30, pp. 831-850, 2008.
[41] Ross, P. J., Taguchi Techniques for Quality Engineering, McGraw-Hill, New York, 1996.
[42] Vapnik, V. N., Statistical Learning Theory, John Wiley &Sons, Inc., New York, 1998.
[43] Vapnyarskii, I. B., “Lagrange Multipliers”, Encyclopedia of Mathematics, Springer, 2001.
[44] Kuhn, H. W., Tucker, A. W., “Nonlinear Programming”, Proceedings of 2nd Berkeley Symposium. Berkeley: University of California Press., pp. 481-492, 1951.
[45] Quinlan, J. R., C4.5: Programs for Machine Learning, Morgan Kaufmann Publishers, San Mateo, CA, 1993.
[46] Specht, D. F., “Probabilistic Neural Network and the Polynomial Adaline as Complementary Techniques for Classification”, IEEE Transactions on Neural Network, Vol. 1, pp. 111-121, 1990.
[47] Guo, L., Huang, D. S., and Zhao, W., “The Optimization of Radial Basis Probabilistic Neural Networks Based on Genetic Algorithms”, IEEE International Conference on Neural Networks, Vol. 4, pp. 3213-3217, 2003.
[48] Burrascano, P., “Learning Vector Quantization for the Probabilistic Neural Networks”, IEEE Transaction on Neural Networks, Vol. 2, pp. 458-461, 1991.