近年來隨著科技進步，高亮度發光二極體(Light Emitting Diode, LED)開始應用在現代生活中，使得高亮度LED需求量漸增，且高亮度LED 具有省電、低溫、不含汞、體積小易於產品設計等優勢，LED 在未來將成為發光源的趨勢，其中LED的品質由電氣特性及表面完整性來決定，因此LED表面檢驗是相當重要的。
本研究開發一套LED自動光學檢測(Automatic Optical Inspection, AOI)系統，應用影像處理技術判斷LED晶粒外觀瑕疵，檢測項目包含破晶、多胞胎、電極區刮傷瑕疵、電極區殘金瑕疵及發光區瑕疵。首先，將影像中晶粒分割出來，本文應用乘冪律轉換(Power-Law Transformation)調整對比明暗度，以統計式(Statistical)門檻值分割前景與背景，接著對獲得的影像進行斷開(Opening)運算以去除不需要的雜點，並保留完整的晶粒外型區塊，並對此區塊做標籤化(Labeling)處理，獲得其外型的特徵值，再由外型特徵值分析晶粒完整性，接著使用索貝爾(Sobel)測邊與哈克轉換(Hough transform)判斷晶粒的角度，並對傾斜角度的晶粒進行角度補償。在晶粒內部結構分析上，先對電極區分析，先利用拉普拉斯(Laplace)轉換銳化(Sharpening)原始影像，再用統計式門檻值取得電極區的特徵值。最後對發光區做分析，將原始影像以直方圖等化(Histogram Equalization)配合乘冪律轉換擴展其灰階度分佈，再利用統計式門檻值取得發光區的特徵值。
綜合所有特徵值以分析LED各部位，在晶粒影像上分析到多種的瑕疵，經過自行研究的檢測系統驗證454顆晶粒影像，總辨識率高達98.24%，証明本研究發展出一套適合於LED瑕疵的檢測系統。

Due to the improving technology these recent years, high brightness light emitting diode (LED) is applied in actual life, and the demand of high brightness light emitting diode is increasing. The advantage of high brightness light emitting diode is the efficient little electricity needed, low temperature, the absence of mercury, the small volume and easy production of the design, so light emitting diode will become the tendency of luminary in the future. The quality of light emitting diode is controlled by electrical properties and surface completeness. Therefore the surface inspection of light emitting diode is very important.
The automatic optical inspection (AOI) system for light emitting diode developed by this research applied image processing technology to determine the appearance defect of light emitting diode crystalline grain. The inspection items include damaged crystalline grain, multi-afterbirth, scratch mark defect of pad area, remained gold defect of pad area and luminesce area defect. It will first separate the light emitting diode crystalline grain from the image and use the power-law transform to adjust contrast and shading value. After that it will use the statistic threshold method to separate the foreground from the background, and then process it with the opening operator which can delete unnecessary noise while reserving the complete shape of the crystalline grain. Then it will process it with labels which can obtain the characteristics of shape, and then analyze the crystalline grain completeness by the characteristics of shape. Next, the Sobel edge inspection and Hough transform is used to determine the angle of crystalline grain. The angle compensation for the applies dip angle crystalline grain. For the pad area analysis, the Laplace transform can sharpen the original image, so the statistic threshold method will obtain the characteristics of the pad area. For the luminesce area analysis, the histogram equalization and power-law transform is used to stretch the gray distribution, and is processed with the statistic threshold method which will obtain the characteristics value of luminesce area.
The automatic optical inspection system concludes all of the characteristics value to analyze each part of light emitting diode, and then analyzes for any kind of defects in the crystalline grain image. Then it investigates 454 grains of the crystalline grain images by self-developed inspection system. The distinguish rate is 98.24%, therefore it is clear that this research has successfully developed a set of defect inspection system applicable to the light emitting diode.

1. A. Kundu and J. Zhou, “Combination Median Filter,” IEEE Trans. on Image Processing, Vol. 1, No. 3, pp. 422-429, 1992.
2. P. S. Windyga, “Fast Impulsive Noise Removal,” IEEE Trans. on Image Processing, Vol. 10, No. 1, pp. 173-179, 2001.
3. C. Dimitry, “Texture Imperfection, Pattern Recognition Letters,” Vol. 6, pp. 45-50 (1987).
4. A. L. Amet and A. Ercil, “Texture Defect Detection Using Domain Co-occurrence Matrices,” Image Analysis and Interpretation, 1998 IEEE Southwest Symposium, pp. 205-210, 1998.
5. H. Sari-Sarraf and S.J. James, “Robust Defect Segmentation in Woven Fabrics,” Proceedings of the 1998 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 938-944.
6. E. N. Malamasa, E. G. M. Petrakisa, M. Zervakisa, L. Petitb, and J. D. Legat, “A Survey on Industrial Vision Systems, Applications and Tools,” Image and Vision Computing, Vol.21, pp. 171-188, 2003.
7. W. A. Bently, “The Inspection: An Automatic Optical Printed Circuit Board(PCB) Inspection,” Proc. Soc. Photo-Optical Instrumn. Engrs., 201, pp. 37-47, 1979.
8. D. Demir, S. Birecik, F. Kurugollu, M. Sezgin, I.O. Bucak, B. Sankur and E. Anarim, “Quality Inspection in PCBs and SMDs Using Computer Vision Techniques,” Industrial Electronics, Control and Instrumentation, 1994.IECON '94., 20th International Conference on Volume: 2 , 1994 , Page(s):857 -861 Vol. 2.
9. E. K. Khwang and D. P. Mital, “A Transputer-based Automated Visual Inspection System for Electronic Devices and PCBs,” Optics and Lasers in Engineering, Vol. 22, pp. 161-180, 1995.
10. E. K. Teoh, D. P. Mital, B. W. Lee and L. K. Wee, “Automated Visual Inspection of Surface Mount PCBs,” IECON '90., 16th Annual Conference of IEEE, Vol.1, pp. 576-580, 1990.
11. H. Yoda, Y. Ohuchi, Y. Taniguchi and M. Ejiri, “An Automatic Wafer Inspection System Using Pipelined Image Progressing Techniques,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 10, No. 1, pp. 4-16 (1988).
12. G. Stanke, T. Wohrle and Schilling, “High-Precision and Versatile Optical Measurement of Different Sides of IC’s in the Confection Process using only one Viewpoint,” F. Industrial Electronics Society, 1998. IECON '98. Proceedings of the 24th Annual Conference of the IEEE Volume: 4, 1998, Page(s): 2425 -2427 vol.4.
13. A. Shapiro,“Automatic Classification of Wafer Defects：Status and Industry Needs,” IEEE Transactions on omponents, Packaging, and Manufacturing Technology, Vol. 20, No. 2, 1997.
14. B. E. Dom and V. Brecher, “Recent Advances in the Automatic Inspection of Integrated Circuits for Pattern Defects,” Machine Vision and Applications, Vol. 8, pp. 5-19, 1995.
15. C. H. C. Jack, L. Maria, C. Wesley, B. David and L. Gray, “A Vision System for Automated Dopant Wafer Inspection,” Vision’89 Conference Proceedings, pp. 5-25~5-41, 1989.
16. M. A. Telles, High Performance Borland C++ Builder, Coriolis Group Books, Albany, N. Y., 1997.
17. 余明興等編著，Borland C++ Builder 6程式設計經典，文魁資訊，2002。
18. 蔡孟凱等編著，C++ Builder 6 完全攻略，金禾資訊，2005。
19. 黃文吉編著，C++ Builder與影像處理，儒林圖書有限公司 ，2002。
20. 蕭進松編著，數位影像處理，全華科技圖書股份有限公司，1999。
21. 鍾國亮編著，影像處理與電腦視覺第三版，東華書局，台北，2006。
22. R. C. Gonzalez and R. E. Woods, Digital Image Processing, Second Edition, Prentice Hall, 2002.
23. I. E. Sobel, “Camera Models and Machine Perception,” Ph.D. thesis, Stanford University, May 1970.
24. N. Otsu, “A Threshold Selection Method from Gray-Level Histogram,” IEEE Trans. Syst., Man, Cybern., Vol. SMC-8, pp. 62-66, 1978.
25. M. Cherist, J. N. Said and C. Y. Suen, “A Recursive Thresholding Technique for Image Segmentation,” IEEE Trans. on Image Processing, Vol. 7, No.6, pp. 918-921, 1998.
26. Y. Du, C. I. Chang and P. D. Thouin, “Unsupervised Approach to Color Video Thresholding,” Optical Engineering, Vol. 43, No. 2, pp. 282-289, 2004.
27. F. Chang and C. J. Chen, “A Component-Labeling Algorithm Using Contour Tracing Technique,” Proc. of Seventh International Conference, Document Analysis and Recognition, pp. 741-745, 2003.
28. R. O. Duda and P. E. Hart, “Random Sample Consensus: A Paradigm for Model Fitting with Application to Image Analysis and Automated Cartography,” Commun. ACM, 24(6), 1981, pp. 381-395.
29. P. V. C. Hough, “Method and Means for Recognizing Complex Patterns,” U. S. Patent 3, 069, 654, Dec. 18, 1962.
30. 吳上立/林宏墩，C語言數位影像處理，全華科技圖書股份有限公司，2006。