臺灣博碩士論文加值系統

本研究涉及保護線性滑軌的密封墊片 （SG） 中的缺陷檢測。對於SG缺陷的生產和品質控制，傳統的檢測方法需要大量的人力。由於視覺疲勞造成的誤差，人類的判斷標準無法標準化。為此，本研究提出了一種基於卷積神經網路（CNN）的深度學習模型。SG數據集用於YOLO模型的高效檢測模型方法。YOLOv8 模型用於判斷五種缺陷類型：毛邊、未分離、劃痕、髒污和斷裂。
在結果本研究驗證所提出的模型在缺陷檢測方面的有效性。YOLOv8 以 97.5% 的準確率和 6.4 秒的平均檢測時間獲得了出色的結果。透過召回率和精確度評估，結果顯示所有模型都具有出色的性能。這項研究凸顯了 YOLO 模型在高效缺陷檢測方面的潛力，為線性滑軌 SG 的簡化品質評估鋪平了道路。

This study addresses defect detection in seal gasket (SG) protecting linear railways. For production and quality control of SG defects, traditional inspection methods need a lot of labor. The evaluation standards of humans cannot be standardized because of errors caused by visual fatigue. For this reason, this study proposed a deep learning model based on a convolutional neural network (CNN). The SG dataset is used in the efficient inspection model approach that uses YOLO models. YOLOv8 model is employed for evaluating performance against five defect types: burr, unseparated, scratched, dirty, and fracture.
Results demonstrated the effectiveness of the proposed models in defect detection. YOLOv8 obtained outstanding results with an accuracy of 97.5% and an average detection time of 6.4 seconds. Recall and precision are also evaluated, revealing strong performance across all models. This study highlights the potential of YOLO models in efficient defect detection, paving the way for stream-lined quality assessment in SG for linear railways.

摘要......i
Abstract......ii
Acknowledgments......iii
Table of Contents......iv
List of Tables......v
List of Figures......vi
Chapter 1 Introduction......1
1.1 Overview Research......1
1.2 Literature Review......2
1.3 Methodology......5
1.4 Thesis Organization......6
Chapter 2 Dataset of the SG......7
2.1 Dataset of SG......7
2.2 Types of SG Defects......10
2.3 Annotation of SG......13
Chapter 3 Training of the inspection models......15
3.1 YOLOv8-based SG model......15
3.2 YOLOv8-based SG model without burr defect......22
Chapter 4 Testing of the inspection models......27
4.1 Testing of YOLOv8-based SG model......27
4.2 Testing of YOLOv8-based SG model without burr defect......30
Chapter 5 Conclusion......33
Reference......35
Extended Abstract......38

[1] Z. Li, F. Liu, W. Yang, S. Peng and J. Zhou, A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects, IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 12, pp. 6999-7019, Dec. 2022.
[2] Z. Li, C. Peng, G. Yu, X. Zhang, Y. Deng, and J. Sun, Light-head r-cnn: In defense of two-stage object detector, Computer Vision and Pattern Recognition, rXiv preprint arXiv:1711.07264, 2017.
[3] X. Sun, X. Jia, Y. Liang, M. Wang, and X. Chi, A Defect Detection Method for a Boiler Inner Wall Based on an Improved YOLO-v5 Net-work and Data Augmentation Technologies, IEEE Access, vol. 10, pp. 93845-93853, 2022, Australia.
[4] X. Luo, Q. Ni, R. Tao, and Y. Shi, A Lightweight Detector Based on Attention Mechanism for Fabric Defect Detection, IEEE Access, vol. 11, pp. 33554-33569, 2023, Australia.
[5] C. Wang, Z. Zhou, and Z. Chen, An Enhanced YOLOv4 Model With Self-Dependent Attentive Fusion and Component Randomized Mosaic Augmentation for Metal Surface Defect Detection, IEEE Access, vol. 10, pp. 97758-97766, 2022, Australia.
[6] Y. Wang, H. Wang, and Z. Xin, Efficient Detection Model of Steel Strip Surface Defects Based on YOLO-V7, IEEE Access, vol. 10, pp. 133936-133944, 2022, Australia.
[7] L. Jiang, B. Yuan, Y. Wang, Y. Ma, J. Du, F. Wang, J. Guo, MA-YOLO: A Method for Detecting Surface Defects of Aluminum Profiles With Attention Guidance, IEEE Access, vol. 11, pp. 71269-71286, 2023, Australia.
[8] W. Chen, Z. Huang, Q. Mu, and Y. Sun, PCB Defect Detection Method Based on Transformer-YOLO, IEEE Access, vol. 10, pp. 129480-129489, 2022, Australia.
[9] Z. Zheng, J. Zhao, and Y. Li, Research on Detecting Bearing-Cover Defects Based on Improved YOLOv3, IEEE Access, vol. 9, pp. 10304-10315, 2021, Australia.
[10] G.Q. Wang, C.Z. Zhang, M.S. Chen, Y.C. Lin, X.H. Tan, P. Liang, Y.X. Kang, W.D. Zeng, and Q. Wang, Yolo-MSAPF: Multiscale Alignment Fusion With Parallel Feature Filtering Model for High Accuracy Weld Defect Detection, IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1-14, 2023, Art no. 5022914, China.
[11] J.W. Zhang, K.C. Lee, and G.A.K. Reddy, Rubber Gasket Defect Classification by VGG16 model, 2022 IEEE 4th Eurasia Conference on IOT, Communication and Engineering (ECICE), pp. 494-499, 2022, Yunlin, Taiwan.
[12] J.J. Liao, J.W. Zhang, B.E. Liu, K.C. Lee, Classification of Guide Rail Block by Xception Model, Applied Functional Materials, vol. 2, pp. 17-24, 2022, Singapore.
[13] B.E. Liu, J.J. Liao, J.W. Zhang, K.C. LEE, 以卷積神經綗路做漳軌塊缺陷的檢測,Journal of Automation Intelligence an Robotics, no. 44, 2023, Taiwan.
[14] T.H. Nguyen, H.L. Nguyen, N.T. Bui, T.H. Bui, V.B. Vu, H.N. Duong, and H.H. Hoang, Vision-Based System for Black Rubber Roller Surface Inspection, Applied Sciences, vol. 13, no. 15, 2023, Switzerland.
[15] C.C. Ho, E. Su, P.C. Li, M.J. Bolger, and H.N. Pan, Machine Vision and Deep Learning Based Rubber Gasket Defect Detection, Advances in Technology Innovation, vol. 5, no.2, pp. 76-83, 2020, Taiwan.
[16] J.W. Zhang, Inspection and Classification of the Defects in Rubber Gaskets, Master Thesis, National Formosa University, 2023, Taiwan.
[17] N. Becherer, J. Pecarina, S. Nykl, and K. Hopkinson, Improving optimization of convolutional neural networks through parameter fine-tuning, Neural Computing and Applications, vol. 31, no.8, pp. 3469-3479, 2019.
[18] Tzutalin, LabelImg, Git code, https://github.com/tzutalin/labelImg, 2015.
[19] G. Jocher, A. Chaurasia, and J. Qiu, YOLO by Ultralytics (Version 8.0.0) Computer software, https://github.com/ultralytics/ultralytics, 2023.