在高價值螺牙生產的過程中，往往會需要對產品進行全檢，以確保每一個螺牙皆符合規定。這種人工檢測方式不僅效率低下，而且需要對檢測人員進行專業培訓。 即便如此，仍然存有人為疏失的問題。為此本論文提出一種基於深度學習與自動光學檢測技術的螺外牙檢測技術，透過視覺辨識技術抓取多角度的螺絲扣件輪廓、並藉由深度殘差網路做特徵萃取後經由一類支援向量機(OCSVM)進行異常值檢測，判斷扣件屬於良品或不良品，藉此實現更加細緻的全檢效果，同時基於ISO965規範評斷，由深度學習模型提出標準化的螺牙健康指標，透過這種檢測機制能使成本有效降低且容易移植，後續進行擴廠或大規模生產也能更加的簡易，在製造過程中能夠代替人工檢測，滿足多數場合的需求，具備健康指標的系統也能在後續螺牙補救的過程中，提供一定程度的依據，判斷螺牙能否被補救，進一步節省人事成本的開銷，提升效率。

In the process of high-value screw production, it is often necessary to conduct a full inspection of the product to ensure that each screw meets the requirements. This manual inspection method is not only inefficient, but also requires professional training for inspectors. Even so, there is still the problem of human error. For this reason, this paper proposes a screw external tooth detection technology based on deep learning and automatic optical detection technology. The multi-angle screw fastener contour is captured by visual recognition technology, and the feature extraction is performed by the deep residual network and then processed by OCSVM(One class support vector machine). Outlier detection, judging whether the fasteners belong to good or bad products, to achieve a more detailed full inspection effect. At the same time, based on the ISO965 standard judgment, a standardized screw tooth health index is proposed by the deep learning model. This detection mechanism can be cost effective. It can be reduced and transplanted easily, and subsequent expansion or mass production can also be simpler. It can replace manual inspection in the manufacturing process and meet the needs of most occasions. The system with health indicators can also be used in the process of subsequent screw tooth recovery. Provide a certain degree of basis to judge whether the screw can be remedied, further save the personnel cost and improve the efficiency.

中文摘要 III
Abstract IV
誌謝 V
目錄 VI
圖目錄 IX
表目錄 XII
名詞縮寫表 XIII
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究流程 4
1.3 論文架構 5
第二章 文獻探討 6
2.1 圖像特徵萃取 6
2.1.1 ROI興趣域(region of interest) 6
2.1.2 圖像平滑化(Blur) 7
2.1.3 二值化(Adaptive Binarization)[9] 10
2.2 波形特徵萃取 11
2.2.1 S-G 濾波器(savitzky-golay filter)[10] 11
2.2.2 離群值偵測(outlier detection)[11] 12
2.3 深度學習模型 13
2.3.1 XGBoost(eXtreme Gradient Boosting)[12] 13
2.3.2 OCSVM(one class support vector machine)[13] 14
2.3.3 1D CNN(1D Convolutional Neural Network)[15] 15
2.3.4殘差收縮網路（Deep Residual Shrinkage Network）[16] 17
2.4 檢測指標與方法 20
2.4.1 二元混淆矩陣(Confusion matrix) 20
2.4.2 ROC (Receiver operating characteristic curve) & AUC ( Area Under Curve) 21
第三章 研究方法與AOI自檢組件設計 23
3.1 AOI機台架構設計 23
3.1.1 鏡頭比較與挑選 23
3.1.2 調光模組規格設計[22] 27
3.1.3 AOI自動光學檢設環境設置[23] 31
3.1.4 調光模組與鏡頭整合 33
3.1.5 機台建置 34
3.2 檢測流程 36
3.2.1 檢測規範與流程 36
3.2.2 數據收集 38
3.3 圖像與波函數處理流程 39
3.3.1 ROI圖像定位 39
3.3.2 二值化(Binarization ) 40
3.3.3 二維圖像波形化 41
3.3.4 波形平滑化(Wave smoothing) 42
3.3.5 波峰波谷萃取(Peak and trough extraction ) 43
3.3.6 計算間距、深度、錐度 45
3.4 模型建立與提升方法 48
3.4.1 基礎1D CNN設計 48
3.4.2 注意力機制與軟化閾值 50
3.4.3 特徵萃取器與強分類模型 54
3.4.4模型訓練與超參數設定 55
第四章 研究結果與討論 57
4.1各階段網路模型比較 57
4.2 損失函數比較 58
4.3 不同強分類器效果 59
第五章 結論與未來展望 61
5.1 結論 61
5.2 未來展望 62
參考文獻 63

[1] R. T. Chin and C. A. Harlow, “Automated Visual Inspection: A Survey.” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 4, no. 6, pp. 557-573, Nov. 1982, doi: 10.1109/TPAMI.1982.4767309.
[2] Mao, Qingzhou, et al. “A rigorous fastener inspection approach for high-speed railway from structured light sensors.” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 143, pp. 249-267, 2018.
[3] Cui, Hao, Qingwu Hu, and Qingzhou Mao. “Real-time geometric parameter measurement of high-speed railway fastener based on point cloud from structured light sensors.” Sensors, vol. 18, no. 11, pp. 3675, Sep 2018, doi: 10.3390/s18113675.
[4] H. Kuo, J. Xu, C. Yu and J. Yan, “Screw defect detection system based on AI image recognition technology.” 2020 International Symposium on Computer, Consumer and Control (IS3C), 2020, pp. 493-496, doi: 10.1109/IS3C50286.2020.00134.
[5] ISO 965 General Purpose Metric Screw Threads – Tolerances – Part 5: Limits of Sizes for Internal Screw Threads to Mate with Hot-Dip Galvanized External Screw Threads with Maximum Size of Tolerance Position H Before Galvanizing International Organization for Standardization, Switzerland (1998).
[6] P. Kovesi, “Fast Almost-Gaussian Filtering.” 2010 International Conference on Digital Image Computing: Techniques and Applications, pp. 121-125, 2010, doi: 10.1109/DICTA.2010.30.
[7] Tao Chen, Kai-Kuang Ma and Li-Hui Chen, “Tri-state median filter for image denoising.” in IEEE Transactions on Image Processing, vol. 8, no. 12, pp. 1834-1838, Dec. 1999, doi: 10.1109/83.806630.
[8] C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images.” Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271), pp. 839-846, 1998, doi: 10.1109/ICCV.1998.710815.
[9] S. Mo, H. Gan, R. Zhang, Y. Yan and X. Liu, “A novel edge detection method based on adaptive threshold.” 2020 IEEE 5th Information Technology and Mechatronics Engineering Conference (ITOEC), pp. 1223-1226, 2020, doi: 10.1109/ITOEC49072.2020.9141577.
[10] Press, W. H., & Teukolsky, S. A. “Savitzky‐Golay smoothing filters.” Computers in Physics, vol. 4, pp. 669-672, doi: 10.1063/1.4822961
[11] HODGE, Victoria; AUSTIN, Jim. “A survey of outlier detection methodologies.” Artificial intelligence review, vol. 22, no. 2, pp. 85-126, 2004.
[12] Chen, T., & Guestrin, C. “Xgboost: A scalable tree boosting system. “ 22nd acm sigkdd international conference on knowledge discovery and data mining. pp. 785-794, Aug 2016.
[13] Schölkopf, et al. “Support vector method for novelty detection. Advances in neural information processing systems.”, NIPS, vol. 12, 1999.
[14] Schölkopf, Bernhard, et al. “Support vector method for novelty detection.” Advances in neural information processing systems, vol. 12 1999.
[15] S. Kiranyaz, T. Ince, O. Abdeljaber, O. Avci and M. Gabbouj, “1-D Convolutional Neural Networks for Signal Processing Applications” ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 8360-8364, doi: 10.1109/ICASSP.2019.8682194.
[16] M. Zhao, S. Zhong, X. Fu, B. Tang and M. Pecht, “Deep Residual Shrinkage Networks for Fault Diagnosis.” in IEEE Transactions on Industrial Informatics, vol. 16, no. 7, pp. 4681-4690, July 2020, doi: 10.1109/TII.2019.2943898.
[17] He, Kaiming, et al. “Deep residual learning for image recognition.” IEEE conference on computer vision and pattern recognition., pp. 770-778, 2016.
[18] D. L. Donoho, “De-noising by soft-thresholding.” in IEEE Transactions on Information Theory, vol. 41, no. 3, pp. 613-627, May 1995, doi: 10.1109/18.382009.
[19] J. Hu, L. Shen and G. Sun, “Squeeze-and-Excitation Networks.” 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 7132-7141, doi: 10.1109/CVPR.2018.00745.
[20] Mandrekar, Jayawant N. “Receiver operating characteristic curve in diagnostic test assessment.” Journal of Thoracic Oncology, vol. 5, no. 9, pp. 1315-1316. 2010.
[21] Lobo, Jorge M., Alberto Jiménez‐Valverde, and Raimundo Real. “AUC: a misleading measure of the performance of predictive distribution models.” Global ecology and Biogeography 17.2 (2008): 145-151
[22] Cheng, Chih-Kai, and Hung-Yin Tsai. “Enhanced detection of diverse defects by developing lighting strategies using multiple light sources based on reinforcement learning.” Journal of Intelligent Manufacturing, pp. 1-13, 2021.
[23] Lu, Rong Sheng, et al. “AOI techniques for surface defect inspection.” Applied mechanics and Materials. Vol. 36. Trans Tech Publications Ltd, 2010.
[24] Wong, Tzu-Tsung. “Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation.” Pattern Recognition 48.9 (2015): 2839-2846.
[25] Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. “Yolov4: Optimal speed and accuracy of object detection.” arXiv preprint arXiv:2004.10934 (2020).