隱形眼鏡作為產品發貨前的檢查很重要，因為隱形眼鏡直接接觸眼睛，隱形眼鏡上的微小表面缺陷可能會導致眼睛不適甚至傷害眼睛。隱形眼鏡上的人工檢查具有缺陷，例如不一致或未能發現主要缺陷。這就需要一種自動光學檢測系統來檢測隱形眼鏡。商用隱形眼鏡存在多種缺陷，其中一些缺陷相對難以檢測。本研究旨在創建一種基於參數的綜合計算機視覺算法，使用 YOLOv4 有效檢測隱形眼鏡中心區域和環區域的缺陷。該算法可以成功地對中心區域和環區域的刮痕、P 點、毛髮和未知物體進行缺陷檢測。該算法以 0.5 閾值置信度得分實現了 mAP 0.90 和 99% 的準確度。檢測隱形眼鏡邊緣區域或邊緣缺陷的演算法的第二部分，其中90%主要使用圖像處理技能。該算法中的主要圖像處理方法包括warp-polar、HSV控制顏色範圍、腐蝕和擴張以找出邊緣缺陷的位置，包括毛刺和裂紋。人工智能還應用於邊緣檢測功能的圓回歸和最小二乘優化等算法。邊緣預處理數據的統計數據顯示為 0.85 AP 和 99% 的準確率。本文提出的整體演算法證明了隱形眼鏡缺陷檢測對高度先進的自動光學檢測系統的價值和潛力。

The inspection on the contact lens before shipping as products is important because the lens contacts the eye directly and small surface defects on the contact lens may cause discomfort on the eye or even injure it. Human inspection on the contact lens is inherited with drawbacks such as inconsistency or failure of finding major defects. This creates a need of an automatic optical inspection system to detect the contact lens. There are many kinds of defects existing in commercial contact lens, some of which are relatively difficult to be detected. This research aims to create a parameter-based comprehensive computer vision algorithm, using YOLOv4 for detecting defects in the center area and the ring area of the contact lens with effectiveness. The algorithm can successfully defect scratches, P-points, hairs, and unknow objects in both the center area and the ring area. The algorithm achieves mAP 0.90 and 99% accuracy with 0.5 threshold confidence score. The second part of the algorithm for detecting defects on the rim area or the edge of a contact lens, in which 90% mainly uses image processing skills. The main image processing methods in this algorithm include warp-polar, control color range by HSV, erode, and dilate to find out the locations of defects on the edge, including burrs and cracks. Artificial intelligence is also applied to the algorithm on such as circle regression and least square optimization for the edge detection function. Statistics for preprocessing data on edge shows 0.85 AP and 99% accuracy. The overall algorithm proposed in this thesis demonstrates the value and potential on the defect detection of the contact lens towards a highly advanced automatic optical inspection system.

REFERENCES
[1] J. Bergin, A technical overview of the development, manufacturing, and future of contact lenses. .
[2] C. Y. Wang, H. Y. Mark Liao, Y. H. Wu, P. Y. Chen, J. W. Hsieh, and I. H. Yeh, “CSPNet: A new backbone that can enhance learning capability of CNN,” IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work., vol. 2020-June, pp. 1571–1580, 2020, doi: 10.1109/CVPRW50498.2020.00203.
[3] J. Yang, S. Li, Z. Wang, H. Dong, J. Wang, and S. Tang, “Using deep learning to detect defects in manufacturing: A comprehensive survey and current challenges,” Materials (Basel)., vol. 13, no. 24, pp. 1–23, 2020, doi: 10.3390/ma13245755.
[4] D. Yang, Y. Cui, Z. Yu, and H. Yuan, “Deep Learning Based Steel Pipe Weld Defect Detection Deep Learning Based Steel Pipe Weld Defect Detection.”
[5] P. M. Bhatt et al., “Image-Based Surface Defect Detection Using Deep Learning: A Review,” J. Comput. Inf. Sci. Eng., vol. 21, no. 4, 2021, doi: 10.1115/1.4049535.
[6] T. Araújo, G. Aresta, A. Galdran, P. Costa, A. M. Mendonça, and A. Campilho, “Uolo - Automatic object detection and segmentation in biomedical images,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 11045 LNCS, pp. 165–173, 2018, doi: 10.1007/978-3-030-00889-5_19.
[7] C.-L. Chang, W.-H. Wu, and C.-C. Hwang, “Automatic optical inspection method for soft contact lenses,” Int. Conf. Opt. Photonic Eng. (icOPEN 2015), vol. 9524, no. July 2015, p. 952402, 2015, doi: 10.1117/12.2182366.
[8] K. M. M. Optometry, “Manufacturing process of contact lens,” no. May, pp. 0–4, 2020.
[9] M. Nadimpalli, “Artificial Intelligence Risks and Benefits,” Int. J. Innov. Res. Sci. Eng. Technol., vol. 6, no. 6, pp. 2–4, 2017.
[10] N. N.J. and A. B.K., “Role of Computer Vision in Automatic Inspection Systems,” Int. J. Comput. Appl., vol. 123, no. 13, pp. 28–31, 2015, doi: 10.5120/ijca2015905603.
[11] C. S. A. Musgrave and F. Fang, “Contact lens materials: A materials science perspective,” Materials (Basel)., vol. 12, no. 2, pp. 1–36, 2019, doi: 10.3390/ma12020261.
[12] A. Hegbom, “Contact lenses,” Tidsskr. Nor. Laegeforen., vol. 119, no. 11, p. 1651, 1999.
[13] M. Rebsamen, J. M. Boucheix, and M. Fayol, “Quality control in the optical industry: From a work analysis of lens inspection to a training programme, an experimental case study,” Appl. Ergon., vol. 41, no. 1, pp. 150–160, 2010, doi: 10.1016/j.apergo.2009.07.004.
[14] M. Chowdhury and A. W. Sadek, “Advantages and Limitations of Artificial Intelligence,” Artif. Intell. Appl. to Crit. Transp. Issues, vol. 6, no. November 2012, pp. 6–8, 2012, [Online]. Available: https://www.researchgate.net/profile/Said_Easa/publication/273576102_Design_and_construction_of_transportation_infrastructure_httponlinepubstrborgonlinepubscircularsec168pdf/links/55097a910cf26ff55f85932b.pdf#page=14.
[15] R. Breton and É. Bossé, “The Cognitive Costs and Benefits of Automation The Benefits of the Automation,” Nato, no. May, pp. 1–12, 2003.
[16] D. Lu and Q. Weng, “A survey of image classification methods and techniques for improving classification performance,” Int. J. Remote Sens., vol. 28, no. 5, pp. 823–870, 2007, doi: 10.1080/01431160600746456.
[17] S. S. Nath, G. Mishra, J. Kar, S. Chakraborty, and N. Dey, “A survey of image classification methods and techniques,” 2014 Int. Conf. Control. Instrumentation, Commun. Comput. Technol. ICCICCT 2014, no. December, pp. 554–557, 2014, doi: 10.1109/ICCICCT.2014.6993023.
[18] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., vol. 2016-Decem, pp. 779–788, 2016, doi: 10.1109/CVPR.2016.91.
[19] L. Jiao et al., “A survey of deep learning-based object detection,” IEEE Access, vol. 7, no. 3, pp. 128837–128868, 2019, doi: 10.1109/ACCESS.2019.2939201.
[20] ISO18369-1:2017, Ophthalmic optics — Contact lenses — Part 1: Vocabulary, classification system and recommendations for labelling specifications. 2017.
[21] H. D. Lin, Y. S. P. Chiu, and S. Y. Hsu, “A visual inspection system for quality control of optical lenses,” Int. J. Phys. Sci., vol. 6, no. 11, pp. 2701–2709, 2011, doi: 10.5897/IJPS11.540.
[22] J. Balzategui, L. Eciolaza, and D. Maestro-Watson, “Anomaly detection and automatic labeling for solar cell quality inspection based on Generative Adversarial Network,” 2021, [Online]. Available: http://arxiv.org/abs/2103.03518.
[23] H. Lin, H. K. Dandage, K. Lin, Y. Lin, and Y. Chen, “Efficient Cell Segmentation from Electroluminescent Images of Single-Crystalline Silicon Photovoltaic Modules and Cell-Based Defect Identification Using Deep Learning with,” 2021.
[24] G. Bradski and A. Kaehler, Learning OpenCV: Computer Vision in C++ with the OpenCV Library, 2nd ed. O’Reilly Media, Inc., 2013.
[25] D. Atikah, “済無No Title No Title No Title,” Angewandte Chemie International Edition, 6(11), 951–952. pp. 29–39, 2017, [Online]. Available: https://www.iso.org/standard/66338.html.
[26] G. Bradski, “The OpenCV Library,” Dr. Dobb’s J. Softw. Tools, 2000.
[27] G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” Proc. - 30th IEEE Conf. Comput. Vis. Pattern Recognition, CVPR 2017, vol. 2017-Janua, pp. 2261–2269, 2017, doi: 10.1109/CVPR.2017.243.
[28] J. Božič, D. Tabernik, and D. Skočaj, “Mixed supervision for surface-defect detection: From weakly to fully supervised learning,” Comput. Ind., vol. 129, 2021, doi: 10.1016/j.compind.2021.103459.