隱形眼鏡包裝的完整性與潔淨度對於產品安全與使用者眼部健康具有重要影響。任何異物、毛屑、黑點或雜壓痕跡等瑕疵若未在出廠前檢出，皆可能導致使用者不適甚至引發眼部感染。現有的品質檢測主要依賴人工目視檢查，不僅效率低且容易受到檢測人員疲勞與主觀判斷影響，導致準確率下降。隨著深度學習與影像處理技術的進步，結合人工智慧的自動化檢測系統有望取代傳統人工檢測，提供高準確率與穩定性的品質管控方案。然而，針對隱形眼鏡包裝異物與瑕疵的專門檢測研究仍屬起步階段，技術尚未普及。
本研究旨在建立一套基於深度學習的隱形眼鏡包裝異物與瑕疵檢測系統，利用 OpenCV 進行影像前處理，並採用 YOLO（You Only Look Once）系列之YOLOv8 及 YOLOv11 模型對瑕疵進行定位與分類，最終將樣品分為良品（合格）或不良品（不合格）。研究範圍涵蓋影像前處理與已標註樣品的模型訓練。本研究透過專門針對隱形眼鏡包裝特性所設計的影像前處理流程，排除背景與非檢測區域的干擾，使 YOLO 模型能聚焦於有效檢測區域內的瑕疵特徵，藉此提升模型的檢測效能。系統設定之檢測目標為不良品檢出率維持在 98% 以上、整體準確率達 95% 以上，並盡可能提高良品檢出率，以驗證本系統具備導入實際生產線的可行性，為隱形眼鏡製造過程的品質保證提供有效的解決方案。

Contact lens packaging defects such as foreign objects, lint, and embossing marks pose risks to user ocular health, yet current inspection relies on error-prone manual examination. We propose a two-stage automated system: the first stage extracts an adaptive annular region of interest (ROI) via Otsu thresholding, RANSAC cir-cular fitting, and Canny edge detection; the second stage employs YOLOv8 and YOLOv11 for defect detection and binary classification. Experiments on 946 pack-aging images show that the preprocessed + YOLOv8 combination achieves 99.3% defective detection rate, 91.1% acceptable detection rate, and 95.9% overall accu-racy, meeting all industrial targets. Adaptive ROI outperforms fixed ROI (99.3% vs. 84.6%), and preprocessing reduces false positives by approximately 51%.

Bj¨orck, ˚A. (2024). Numerical Methods for Least Squares Problems. Society for Indus-trial and Applied Mathematics (SIAM), Philadelphia, PA, 2nd edition.
Bochkovskiy, A., Wang, C.-Y., and Liao, H.-Y. M. (2020). YOLOv4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.
Canny, J. (1986). A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 8(6):679–698.
Chernov, N. and Lesort, C. (2005). Least squares fitting of circles. Journal of Mathe-matical Imaging and Vision, 23(3):239–252.
Coope, I. D. (1993). Circle fitting by linear and nonlinear least squares. Journal of Optimization Theory and Applications, 76(2):381–388.
Davies, E. R. (2018). Computer Vision: Principles, Algorithms, Applications, Learning. Academic Press, London, UK, 5th edition.
Fischler, M. A. and Bolles, R. C. (1981). Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Com-munications of the ACM, 24(6):381–395.
Gonzalez, R. C. and Woods, R. E. (2018). Digital Image Processing. Pearson, New York, NY, 4th edition.
Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press, Cambridge, MA.
Haralick, R. M., Sternberg, S. R., and Zhuang, X. (1987). Image analysis using math-ematical morphology. IEEE Transactions on Pattern Analysis and Machine Intelli-gence, 9(4):532–550.
Hartley, R. and Zisserman, A. (2004). Multiple View Geometry in Computer Vision. Cambridge University Press, Cambridge, UK, 2nd edition.
Jocher, G. (2020). YOLOv5. GitHub repository.
Jocher, G., Chaurasia, A., and Qiu, J. (2023). Ultralytics YOLO (version 8.0.0). Com-puter software.
Jocher, G. and Qiu, J. (2024). Ultralytics YOLO11. Computer software.
Kaehler, A. and Bradski, G. (2016). Learning OpenCV 3: Computer Vision in C++ with the OpenCV Library. O’Reilly Media, Sebastopol, CA.
Khanam, R. and Hussain, M. (2024). YOLOv11: An overview of the key architectural enhancements. arXiv preprint arXiv:2410.17725.
Lohr, S. L. (2021). Sampling: Design and Analysis. CRC Press, Boca Raton, FL, 3rd edition.
Malamas, E. N., Petrakis, E. G., Zervakis, M., Petit, L., and Legat, J. D. (2003). A survey on industrial vision systems, applications and tools. Image and Vision Com-puting, 21(2):171–188.
Nixon, M. S. and Aguado, A. S. (2019). Feature Extraction and Image Processing for Computer Vision. Academic Press, London, UK, 4th edition.
Otsu, N. (1979). A threshold selection method from gray-level histograms. IEEE Trans-actions on Systems, Man, and Cybernetics, 9(1):62–66.
Raguram, R., Frahm, J. M., and Pollefeys, M. (2008). A comparative analysis of ransac techniques leading to adaptive real-time random sample consensus. In Proceedings of European Conference on Computer Vision, pages 500–513.
Redmon, J., Divvala, S., Girshick, R., and Farhadi, A. (2016). You only look once: Uni-fied, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 779–788.
Redmon, J. and Farhadi, A. (2017). YOLO9000: Better, faster, stronger. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 6517–6525.
Redmon, J. and Farhadi, A. (2018). YOLOv3: An incremental improvement. arXiv preprint arXiv:1804.02767.
Rosenfeld, A. (1982). Non-maximum suppression of gradient magnitudes makes them easier to threshold. Pattern Recognition Letters, 1:93–94.
Sezgin, M. and Sankur, B. (2004). Survey over image thresholding techniques and quantitative performance evaluation. Journal of Electronic Imaging, 13(1):146–168.
Sobel, I. and Feldman, G. (1968). A 3×3 isotropic gradient operator for image process-ing. In Stanford Artificial Intelligence Laboratory.
Soille, P. (2004). Morphological Image Analysis: Principles and Applications. Springer, Berlin, Germany, 2nd edition.
Steger, C., Ulrich, M., and Wiedemann, C. (2018). Machine Vision Algorithms and Applications. Wiley-VCH, Weinheim, Germany, 2nd edition.
Szeliski, R. (2022). Computer Vision: Algorithms and Applications. Texts in Computer Science. Springer, Cham, Switzerland, 2nd edition.
Yaseen, M. (2024). What is YOLOv8: An in-depth exploration of the internal features of the next-generation object detector. arXiv preprint arXiv:2408.15857.
Zhang, Z. (1997). Parameter estimation techniques: A tutorial with application to conic fitting. Image and Vision Computing, 15(1):59–76.