瑕疵檢測已在各行業各業當中扮演著重要角色，其影響無法被忽視，生產過程中可透過自動光學檢測(Automated Optical Inspection，AOI)系統來判讀是否有瑕疵與區分瑕疵的種類，以確保生產出來的產品是合乎品質要求的。這不單單可以降低產品的不良率、提升產品品質，更可以達到節省成本以及增加品牌商譽，在競爭激烈的市場中取得領先優勢。
本研究旨在降低AOI系統判讀瑕疵時的重複性問題，一般在AOI系統辨識的過程中，會因為不同光線、角度與反射等環境差異下產生不同的判讀結果，輕則列為次級品，重則報廢而無法使用，為避免機器判讀錯誤而花費更多人力成本進行二次檢驗，顯然不符合效益；故本研究將會利用影像辨識演化法與深度學習等相關技術，運用個案公司生產時所累積瑕疵圖片資料庫，來找出重複被判讀的瑕疵。
實驗共分四個階段：第一階段將會針對收集來共12種不同類別的瑕疵圖片，透過K-近鄰演算法(K-Nearest Neighbor，KNN)的分類，來過濾掉圖片中含有雜訊的部分，每個類別各取出20張圖片；第二階段針對圖片進行資料增生，對每張圖片進行平移、縮放、翻轉以及隨機裁剪來模擬不同位置的瑕疵照片；第三階段分別採用尺度不變特徵轉換(Scale-Invariant Feature Transform，SIFT)特徵提取演算法搭配快速近似最近鄰居搜尋(Fast Library for Approximate Nearest Neighbors，FLANN)進行特徵匹配，並與孿生神經網路(Siamese Neural Networks，SNN)各別計算圖片間的相似度；第四階段為將兩種實驗方式的成效互相做對比，最終在最好的情況下，可以達到準確率0.9500、精確率1、召回率0.9000、F1-score為 0.9474的結果。

This research addresses the challenge of repeatability in defect interpretation by Automated Optical Inspection systems. In the recognition process of optical inspection systems, variations in environmental conditions, such as lighting, angles, and reflections, lead to diverse interpretation results. This variability can result in the misclassification of items as substandard or even render them unusable, necessitating additional human inspection, which is evidently not cost-effective. To mitigate these issues, the study employs image recognition methodologies, including evolutionary algorithms and deep learning. The approach leverages a database of defective images accumulated during the production process of a specific company to identify redundantly interpreted defects.
The experiment unfolds in four stages：
In the first stage, using a KNN Classifier to classify collected images into 12 different defect categories, effectively filtering out the noises. Then each categories were chosen 20 images for second stage. The second stage involves data augmentation on images, simulating defects in different positions through translation, scaling, flipping, and random cropping. The third stage utilizes the SIFT for feature extraction, coupled with the FLANN for feature matching. Additionally, SNN are employed to calculate the similarity between images. In the fourth stage, the effectiveness of the two experimental approaches is compared. Ultimately, in the best case, the SNN model achieved results of Accuracy 0.9500, Precision 1, Recall 0.9000, and F1-score 0.9474.

1. 陳均富, 基於賈伯濾波器之紋理分割, in 電機工程學系碩博士班. 2010, 國立成功大學: 台南市. p. 89.
2. American Chemistry Council, I. Center for the Polyurethanes Industry (CPI) Applications & Benefits. [cited 2023 Dec 04]; Available from: https://www.americanchemistry.com/industry-groups/center-for-the-polyurethanes-industry-cpi/applications-benefits.
3. Bennamoun, M. and A. Bodnarova. Automatic visual inspection and flaw detection in textile materials: past, present and future. in SMC'98 Conference Proceedings. 1998 IEEE International Conference on Systems, Man, and Cybernetics (Cat. No.98CH36218). 1998.
4. Bentley, J.L., Multidimensional binary search trees used for associative searching. Communications of the ACM, 1975. 18(9): p. 509-517.
5. Bhatt, P.M., et al., Image-Based Surface Defect Detection Using Deep Learning: A Review. Journal of Computing and Information Science in Engineering, 2021. 21(4).
6. Borwankar, R. and R. Ludwig, An Optical Surface Inspection and Automatic Classification Technique Using the Rotated Wavelet Transform. IEEE Transactions on Instrumentation and Measurement, 2018. 67(3): p. 690-697.
7. Bradley, A.P., The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognition, 1997. 30(7): p. 1145-1159.
8. Bromley, J., et al., Signature Verification using a "Siamese" Time Delay Neural Network. International Journal of Pattern Recognition and Artificial Intelligence, 1993. 7: p. 25.
9. Chen, Y., et al., Surface defect detection methods for industrial products: A review. Applied Sciences, 2021. 11(16): p. 7657.
10. Chi-Ho, C. and G.K.H. Pang, Fabric defect detection by Fourier analysis. IEEE Transactions on Industry Applications, 2000. 36(5): p. 1267-1276.
11. Chopra, S., R. Hadsell, and Y. LeCun. Learning a similarity metric discriminatively, with application to face verification. in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05). 2005.
12. Covestro AG. Texin® for everything between rubber and plastics. [cited 2023 Dec 04]; Available from: https://solutions.covestro.com/en/brands/texin#brandbenefits.
13. Czimmermann, T., et al., Visual-Based Defect Detection and Classification Approaches for Industrial Applications—A SURVEY. Sensors, 2020. 20(5): p. 1459.
14. Ding, S., Z. Liu, and C. Li. AdaBoost learning for fabric defect detection based on HOG and SVM. in 2011 International Conference on Multimedia Technology. 2011.
15. Fawcett, T., An introduction to ROC analysis. Pattern Recognition Letters, 2006. 27(8): p. 861-874.
16. Friedman, J.H., J.L. Bentley, and R.A. Finkel, An algorithm for finding best matches in logarithmic expected time. ACM Transactions on Mathematical Software (TOMS), 1977. 3(3): p. 209-226.
17. Hu, G.-H., Automated defect detection in textured surfaces using optimal elliptical Gabor filters. Optik, 2015. 126(14): p. 1331-1340.
18. Hui-Fuang, N. Automatic thresholding for defect detection. in Third International Conference on Image and Graphics (ICIG'04). 2004.
19. Jiahan, C. and K.J. Anil. A Structural Approach To Identify Defects In Textured Images. in Proceedings of the 1988 IEEE International Conference on Systems, Man, and Cybernetics. 1988.
20. Jung, A.B., et al. imgaug. [cited 2023 Dec 12]; Available from: https://github.com/aleju/imgaug.
21. Kim, D.-H., et al., Smart Machining Process Using Machine Learning: A Review and Perspective on Machining Industry. International Journal of Precision Engineering and Manufacturing-Green Technology, 2018. 5: p. 555-568.
22. Koch, G., R. Zemel, and R. Salakhutdinov. Siamese neural networks for one-shot image recognition. in ICML deep learning workshop. 2015. Lille.
23. Liao, H.C., et al. Guidelines of Automated Optical Inspection (AOI) System Development. in 2018 IEEE 3rd International Conference on Signal and Image Processing (ICSIP). 2018.
24. Lowe, D., Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, 2004. 60: p. 91.
25. Mei, S., Y. Wang, and G. Wen, Automatic Fabric Defect Detection with a Multi-Scale Convolutional Denoising Autoencoder Network Model. Sensors, 2018. 18(4): p. 1064.
26. Melekhov, I., J. Kannala, and E. Rahtu. Siamese network features for image matching. in 2016 23rd International Conference on Pattern Recognition (ICPR). 2016.
27. Muja, M. and D.G. Lowe, Fast approximate nearest neighbors with automatic algorithm configuration. VISAPP (1), 2009. 2(331-340): p. 2.
28. Muja, M. and D.G. Lowe. Fast Matching of Binary Features. in 2012 Ninth Conference on Computer and Robot Vision. 2012.
29. Muja, M. and D.G. Lowe. Flann, fast library for approximate nearest neighbors. in International Conference on Computer Vision Theory and Applications (VISAPP’09). 2009. INSTICC Press Setúbal, Portugal.
30. Muja, M. and D.G. Lowe, Scalable Nearest Neighbor Algorithms for High Dimensional Data. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014. 36(11): p. 2227-2240.
31. Newman, T.S. and A.K. Jain, A Survey of Automated Visual Inspection. Computer Vision and Image Understanding, 1995. 61(2): p. 231-262.
32. Perkins, N.J. and E.F. Schisterman, The inconsistency of "optimal" cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol, 2006. 163(7): p. 670-5.
33. Provost, F. and T. Fawcett, Robust classification for imprecise environments. Machine learning, 2001. 42: p. 203-231.
34. Sakhare, K., et al. Spectral and spatial domain approach for fabric defect detection and classification. in 2015 International Conference on Industrial Instrumentation and Control (ICIC). 2015.
35. Sheu, R.-K., et al. Apparatus and Method of Defect Detection for Resin Films. Applied Sciences, 2020. 10, DOI: 10.3390/app10041206.
36. Silpa-Anan, C. and R. Hartley. Optimised KD-trees for fast image descriptor matching. in 2008 IEEE Conference on Computer Vision and Pattern Recognition. 2008.
37. Tao, X., et al., Automatic Metallic Surface Defect Detection and Recognition with Convolutional Neural Networks. Applied Sciences, 2018. 8(9): p. 1575.
38. Wang, S., Z. Guo, and Y. Liu, An Image Matching Method Based on SIFT Feature Extraction and FLANN Search Algorithm Improvement. Journal of Physics: Conference Series, 2021. 2037(1): p. 012122.
39. Wang, T., et al., A fast and robust convolutional neural network-based defect detection model in product quality control. The International Journal of Advanced Manufacturing Technology, 2018. 94(9): p. 3465-3471.
40. Wu, X., et al. PCB Defect Detection Using Deep Learning Methods. in 2021 IEEE 24th International Conference on Computer Supported Cooperative Work in Design (CSCWD). 2021.
41. Youden, W.J., Index for rating diagnostic tests. Cancer, 1950. 3(1): p. 32-35.
42. Yuan, X.-c., L.-s. Wu, and Q. Peng, An improved Otsu method using the weighted object variance for defect detection. Applied Surface Science, 2015. 349: p. 472-484.
43. Zadrozny, B. and C. Elkan. Obtaining calibrated probability estimates from decision trees and naive bayesian classifiers. in Icml. 2001.
44. Zagoruyko, S. and N. Komodakis Learning to Compare Image Patches via Convolutional Neural Networks. 2015. arXiv:1504.03641 DOI: 10.48550/arXiv.1504.03641.
45. Zhong, B. and Y. Li. Image Feature Point Matching Based on Improved SIFT Algorithm. in 2019 IEEE 4th International Conference on Image, Vision and Computing (ICIVC). 2019.