AOI自動外觀檢驗廣泛的應用於各種精密且產量需求大的半導體產業上，目前半導體製程都是以批量為生產單位，每批產量都介於數十萬顆至數百萬之間，AOI自動外觀檢測若因參數設定不當而造成檢測缺陷逃脫，導致退貨重工，不僅成本上升，嚴重流至客戶端，則須賠償產品損失。
本文首先以分析缺金缺陷的特徵，包括其剖面圖、粗糙度、以及表面色碼分析，以評估光源的影響，再應用田口實驗方法設定量產參數，針對影響成像差異性最顯著的因子：光源種類、亮度、以及照射的角度，進行優化。結果可發現在R、G、B的光源中以綠光最能使銅面及金面二值化的影像呈現明顯差異；在亮度的選擇上以375毫安培的電流為最佳使用參數，照射的角度則以90度最能呈現缺金缺陷的差異。
總體改善成果，經半年的統計數據觀察此次實驗結果，退貨率從原本的7批/月(單批產品生產量為1件)降低為0.5批/月，平均單批作業時間減少了4.93%，而每月耗材費用也降低了95.54%。

AOI (Automated Optical Inspection) is a widely utilized inspection technique in semiconductor industry. Electronic products are getting smaller, thinner and lighter by the evolution of semiconductor technology. The number of substrate unit in each batch also increases. Thus, keeping the quality of products is essential. However, defect loss can occur through the inspection by the AOI machine. Defect loss not only increases the rework time but also promotes the risk of passing abnormal products to customers.
To improve the quality of AOI for substrate with Au defects, the differences in surface color codes between copper and gold were investigated first. Then the area of the deep color block in copper surface and shallow color block in gold surface was counted by using the program ImageJ. To improve the AOI results, the original halogen lamp light source was replaced by LEDs with three types of light (R, G, B). Moreover, the Taguchi method was used to obtain the optimum combination of AOI inspection parameters obtained from the practical manufacturing process. The experimental results show that the color code and color block area are useful for improving the defect detection. The green LED light source reveals a significant image color difference between the copper and gold metal trace. In addition, the condition of using light power with 375mA current and irradiation of 90 degrees angle can result in the maximum image difference between copper and gold metal trace in gray scale.
With the above AOI parameters and 6 months practical observation, the rework times decreases greatly from 7 times per month to 0 times per month; the material cost decreases by 26.79%, and average working time per batch reduces by 4.83%.

[1] Wei-Chien Wang, Shang-Liang Chen, Liang-Bi Chen, Wan-Jung Chang (2017), A MAchine Vision Based AutomAtic Optical Inspection System for Measuring Drilling Quality of Printed Circuit Boards, Access IEEE, vol. 5, pp. 10817-10833.
[2] 林晉宏，應用田口方法品質分析於積層陶瓷電容器之外觀檢驗，碩士論文，工業管理系所，國立屏東科技大學，2012年
[3] 林綉莉，智慧型影像量測與分析之研究與應用於自動化光學檢測，碩士論文，電機工程系碩士班，聖約翰科技大學，2009年
[4] R. S. Lu et al.(2010). AOI Techniques for Surface Defect Inspection, Applied Mechanics and MAterials, Vol. 36, 297-302.
[5] Shuang Li, Zhongsheng Wang (2009) (Northwestern Polytechnical University,Xi’an 710072,China);Aero-engine Rotor Crack Identification System Based on MAchine Vision[J];Computer Measurement & Control
[6] 顏睿余，高性能DSPBased影像擷取辨識系統，碩士論文，電機工程學系研究所，國立中山大學，2001年
[7] Hao Wu, Xianmin Zhang , Hongwei Xie , Yongcong Kuang ,Gaofei Ouyang (2013) , Classification of Solder Joint Using Feature Selection Based on Bayes and Support Vector MAchine, IEEE Transactions on Components, Packaging and MAnufacturing Technology,Volume: 3, Issue: 3
[8] D. B. Perng, Y. C. Chen, M. K. Lee (2005). A novel AOI system for OLED panel inspection, Journal of Physics: Conference Series, Vol.13.
[9] 鄭睿夫，影像處理技術於載帶規範檢測之應用，碩士論文，工程科學系專班，國立成功大學，2002年
[10] 羅光驊，基於RGB色彩分離之夜視鏡顯像良率分析，碩士論文，理工學院工程科技碩士在職專班資訊系統組，國立金門大學，2016年
[11] Fupei Wu, Xianmin Zhang (2011) ,Feature-Extraction-Based Inspection Algorithm for IC Solder Joints,IEEE Transactions on Components, Packaging and MAnufacturing Technology , vol , 1 , Issue,5, pp.689–694
[12] H. Z. Hel-Or and B. A. Wandell (2002), "Object-based illumination classification", Pattern Recognition, Vol. 35, N0.8, August, pp. 1723-1732.
[13] H. Michael (1997), "Elimination of highlights using RGB color distribution and imAge position", Proceedings of the Color ImAging Conference, p205~208.
[14] Chin-Teng Lin, Kan-Wei Fan, and Wen-Chang Cheng, (2005) "An illumination estimAtion scheme for color constancy based on chromAticity histogram and neural network", 2005 International Conference on Systems, MAn and Cybernetics, p 2488-2494.
[15] Hyung Tae Kim, Seung Taek Kim, Young June Cho (2015). Quick and efficient light control for conventional automAtic optical inspection (AOI) systems, International Journal of Precision Engineering and MAnufacturing, Vol.16, 247.
[16] Zerefos, C. S., Tetsis, P., Kazantzidis, A., Amiridis, V., Zerefos, S. C., Luterbacher, J., Eleftheratos, K., Gerasopoulos, E., Kazadzis, S., and Papayannis (2014). A.: Further evidence of important environmental informAtion content in red-to-green ratios as depicted in paintings by great mAsters, Atmos. Chem. Phys., 14, 2987-3015.
[17] Dongsu Lee, Kihak Lee, Kyunghan Chun, Bonghwan Kim (2018). Advanced Science Letters, Vol. 24, 4936-4941.
[18] Jincheng Cheng, Aimin Wang, Yan Ge (2017). The Technology of Quality Statistical Process Control Based on AOI Inspection Results, IEEE.
[19] Guangming Ni; Lin Liu; Jing Zhang; Juanxiu Liu; Yong Liu (2018). High-resolution imAging optomechatronics for precise liquid crystal display module bonding automAted optical inspection. J. of Electronic ImAging, 27(1), 013013.
[20] Bo Peng, Deguang Zheng and Yue Yu (2017). Method to remove wafer surface particles.Journal of Semiconductors, Volume 38, Number 9
[21] Ming-Fu Chen, Chih-Wen Chen, Chih-Yen Chen, Chi-Hung Hwang, Liang-Yin Hwang (2016). An AOI system development for inspecting defects on 6 surfaces of chips.
IEEE International Instrumentation and Measurement Technology Conference Proceedings.
[22] Teng-da Zhang, Rong-sheng Lu (2016). Surface defect inspection of TFT-LCD panels based on 1D Fourier method. Proc. SPIE 9903, Seventh International Symposium on Precision Mechanical Measurements, 990308.
[23] H. T. Young, H. S. Chiou, W. R. Huang (2016).AOI strategies for defects identification of micro-electro mechanical system elements. Advances in MAterials and Processing Technologies. Volume 2 - Issue 2
[24] 潘亨霖，利用色差做Logo辨認，碩士論文，電子工程系，國立臺灣科技大學，2006年
[25] 劉益興，吸收式生化光電檢測技術之研究，碩士論文，資訊電機工程碩士在職專班，逢甲大學，2006年
[26] 郭俊緯，液晶顯示面板之自動化光學檢測系統，碩士論文，機械工程系，國立臺灣科技大學，2006年
[27] 黎榮富，以DOE手法降低SMT光學檢測設備檢測誤判率，碩士論文，管理學院工業工程與管理EMBA專班，國立臺北科技大學，2017年
[28] 林佑任，電路板零件及焊點之自動光學檢測系統，碩士論文，電子工程系，南臺科技大學電子工程系，2018年
[29] 周千惠，彩色OLED亮度檢測及補償機制之研究，碩士論文，自動化及控制研究所，國立臺灣科技大學，2006年
[30] 徐錦淵，高亮度LED光電熱特性量測與分析，碩士論文，平面顯示技術碩士學位學程，國立交通大學，2010年
[31] 邱建智，彩色濾光片瑕疵自動化檢測之輝度值與RGB色度關係之研究，碩士論文，機械工程學系碩士班，中華大學，2010年
[32] 楊政儒，光源照度與RGB灰階理論對彩色光彈應力條紋之影響研究，碩士論文，土木工程學系碩士班，中華大學，2006年