近來由於物聯網、大數據、人工智慧、5G、邊緣及雲端計算等新興資通訊技術快速發展，帶動工業製造場域的智能化。以現場設施的人工巡檢作業為例，透過擴增實境介面，以人工智慧協助辨識物件姿態，偵測裝置可能的異常狀態，啟動對應處理程序，可減少巡檢作業的工作負荷，降低過程中可能的人為疏失。基於此一概念，本研究以流體球閥開關為例，基於深度學習模型，發展偵測狀態異常的智能化工具。此外，分析不同條件下的部署方式，基於模糊運算輔助其規劃決策，考量計算模型（二維影像、三維姿態）、運算裝置（手持式裝置、邊緣裝置與雲端平台）、網路傳輸能力與介面裝置（手持式、頭戴式裝置）的組合影響；並根據不同產業特性，反映於計算時間、預測準確度與建置成本的需求，以及作業現場使用條件限制，建議可行之部署規劃組合；最後透過真實環境的部署與測試結果，驗證研究概念的可行性，作為導入擴增實境智能化應用的參考依據。

The rapid development of information and communication technologies, such as the Internet of Things, big data, artificial intelligence (AI), 5G, edge computing, and cloud computing lead to intelligentization in the manufacturing industry. For example, in manual inspection of on-site facilities, AI can assist human operator to detect anomaly of devices by recognizing object posture and to initiate the troubleshooting procedure via AR interfaces. The workload of the manual inspection and potential human errors are thus significantly reduced. Therefore, this work develops an intelligent solution based on deep learning models for automatic anomaly detection of ball valve switches. In addition, we analyze various factors that affect on-site deployment of the solution, including information types (2D image, 3D pose), computing devices (handheld, edge, cloud), network transmission, and interfacing devices (smartphone, AR goggle). A decision support tool based on fuzzy logics is constructed to assist the deployment planning. The support tool analyzes how the requirement of computational time, prediction precision, and implementation costs change with the characteristics of various industries. Based on the analysis result, it suggests feasible deployment plans subject to limitations of on-site conditions. Finally, test results obtained from real settings demonstrates the effectiveness of the support tool. This work provides valuable guidelines for implementing AR-based intelligent solutions.

摘要 ------------II
Abstract ------------III
目錄 ------------IV
圖目錄 ------------VI
表目錄 ------------VIII
第一章、 緒論 ------------1
1.1 研究背景 ------------1
1.2 研究目的 ------------2
第二章、 文獻回顧 ------------4
2.1 工業物件辨識 ------------4
2.2 基於物件三維姿態估算及應用 ------------5
2.3 智能化方案的實際部署應用 ------------8
2.4 小結 ------------9
第三章、 系統介紹 ------------10
3.1 測試物件說明 ------------10
3.2系統功能 ------------10
第四章、 研究方法 ------------13
4.1 深度學習 ------------13
4.1.1 二維深度卷積模型 ------------14
4.1.2 三維深度卷積模型 ------------15
4.2 建立球閥資料集 ------------17
4.2.1 真實資料集 ------------17
4.2.2 合成資料集 ------------19
4.3 資料傳輸 ------------24
4.4 影像處理與選擇 ------------25
4.5 開關狀態判斷方法 ------------26
第五章、 系統實作 ------------29
5.1 智能化人工輔助巡檢原型系統建置 ------------29
5.2 球閥開關狀態偵測系統結果 ------------35
5.3 系統失效狀況 ------------40
第六章、 系統部署規劃 ------------42
6.1 評估方法 ------------42
6.2 決策算法設計 ------------42
6.3 實際部署規劃 ------------49
6.4 部署方案建議 ------------52
6.5 小結 ------------56
第七章、 結論與未來展望 ------------57
參考文獻 ------------60
附錄一、資通訊整合方案組合 ------------63
附錄二、準確度計算之部份測試資料 ------------64
附錄三、資通訊整合方案量化數值總和 ------------65
附錄四、球閥基底部份合成訓練資料 ------------66
附錄五、球閥握把部份合成訓練資料 ------------68
附錄六、球閥部份真實訓練資料 ------------70

1. Correll, N., Bekris, K. E., Berenson, D., Brock, O., Causo, A., Hauser, K., ... & Wurman, P. R. (2016). Analysis and observations from the first amazon picking challenge. IEEE Transactions on Automation Science and Engineering, 15(1), 172-188.
2. 黃紹源、黃久銘、張凱強 (2021)，擴增實境中基於深度學習之工件三維姿態估算，清華大學工業工程與工程管理學系，大四專題報告。
3. Jiang, Q., Tan, D., Li, Y., Ji, S., Cai, C., & Zheng, Q. (2020). Object detection and classification of metal polishing shaft surface defects based on convolutional neural network deep learning. Applied Sciences, 10(1), 87.
4. Saeed, F., Ahmed, M. J., Gul, M. J., Hong, K. J., Paul, A., & Kavitha, M. S. (2021). A robust approach for industrial small-object detection using an improved faster regional convolutional neural network. Scientific reports, 11(1), 1-13.
5. Mou, X., Cui, J., Yin, H., & Zhou, X. (2019, November). Tracking position and status of electric control switches based on YOLO detector. In International Conference on Intelligent Data Engineering and Automated Learning (pp. 184-194). Springer, Cham.
6. Li, H., Zheng, B., Sun, X., & Zhao, Y. (2017, October). CNN-based model for pose detection of industrial PCB. In 2017 10th International Conference on Intelligent Computation Technology and Automation (ICICTA) (pp. 390-393). IEEE.
7. Kim, Y. H., & Lee, K. H. (2019). Pose initialization method of mixed reality system for inspection using convolutional neural network. Journal of Advanced Mechanical Design, Systems, and Manufacturing, 13(5), JAMDSM0093-JAMDSM0093.
8. Sun, Y., Kantareddy, S. N. R., Siegel, J., Armengol-Urpi, A., Wu, X., Wang, H., & Sarma, S. (2019, October). Towards industrial iot-ar systems using deep learn-ing-based object pose estimation. In 2019 IEEE 38th International Performance Computing and Communications Conference (IPCCC) (pp. 1-8). IEEE.
9. Li, Y., Mo, K., Shao, L., Sung, M., & Guibas, L. (2020, August). Learning 3d part assembly from a single image. In European Conference on Computer Vision (pp. 664-682). Springer, Cham.
10. Wang L., Wang J., Jiao S., Wang M., & Li, S. (2021). Fully convolutional net-work-based registration for augmented assembly systems, to appear, Journal of Manufacturing Systems.
11. Angel, N. A., Ravindran, D., Vincent, P. M., Srinivasan, K., & Hu, Y. C. (2022). Recent Advances in Evolving Computing Paradigms： Cloud, Edge, and Fog Technologies. Sensors, 22(1), 196.
12. Premsankar, G., Di Francesco, M., & Taleb, T. (2018). Edge computing for the Internet of Things： A case study. IEEE Internet of Things Journal, 5(2), 1275-1284.
13. Liu, C., Cao, Y., Luo, Y., Chen, G., Vokkarane, V., Yunsheng, M., ... & Hou, P. (2017). A new deep learning-based food recognition system for dietary assessment on an edge computing service infrastructure. IEEE Transactions on Services Computing, 11(2), 249-261.
14. Skousen, P. L. (2011). Valve handbook. McGraw-Hill Education.
15. xIkwuegbu, I. (2021). AlphaGo Design Principle.
16. Adamopoulou, E., & Moussiades, L. (2020, June). An overview of chatbot tech-nology. In IFIP International Conference on Artificial Intelligence Applications and Innovations (pp. 373-383). Springer, Cham.
17. He, Z., Feng, W., Zhao, X., & Lv, Y. (2020). 6D pose estimation of objects： Recent technologies and challenges. Applied Sciences, 11(1), 228.
18. Kim, S. H., & Hwang, Y. (2021). A Survey on Deep Learning Based Methods and Datasets for Monocular 3D Object Detection. Electronics, 10(4), 517.
19. Hartley, R., & Zisserman, A. (2003). Multiple view geometry in computer vision. Cambridge university press.
20. YOLOv4-tiny, https://github.com/AlexeyAB/darknet
21. Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. M. (2020). Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.
22. YOLOv4-tiny架構, https://www.twblogs.net/a/5f0294f1d496dddbb54259af
23. Wang, C. Y., Liao, H. Y. M., Wu, Y. H., Chen, P. Y., Hsieh, J. W., & Yeh, I. H. (2020). CSPNet: A new backbone that can enhance learning capability of CNN. In Pro-ceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops (pp. 390-391).
24. Tekin, B., Sinha, S. N., & Fua, P. (2018). Real-time seamless single shot 6d object pose prediction. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 292-301).
25. ImageNet, http://www.image-net.org/
26. MS-COCO, http://cocodataset.org/#home
27. PASCAL VOC, http://host.robots.ox.ac.uk/pascal/VOC/
28. Hinterstoisser, S., Lepetit, V., Ilic, S., Holzer, S., Bradski, G., Konolige, K., & Navab, N. (2012, November). Model based training, detection and pose estimation of texture-less 3d objects in heavily cluttered scenes. In Asian conference on computer vision (pp. 548-562). Springer, Berlin, Heidelberg.
29. Brachmann, E., Michel, F., Krull, A., Ying Yang, M., & Gumhold, S. (2016). Uncertainty-driven 6d pose estimation of objects and scenes from a single rgb image. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3364-3372).
30. Xiang, Y., Schmidt, T., Narayanan, V., Fox, D.： PoseCNN： A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes. Robotics： Science and Systems Conference (2018)
31. Saravanan, R., & Sujatha, P. (2018, June). A state of art techniques on machine learning algorithms: a perspective of supervised learning approaches in data classification. In 2018 Second International Conference on Intelligent Computing and Control Systems (ICICCS) (pp. 945-949). IEEE.
32. labelImg, https://github.com/tzutalin/labelImg
33. 周庭 (2020)，基於深度卷積網路之擴增實境鞋品虛擬試穿，清華大學工業工程與工程管理學系，碩士論文。
34. To, T., Tremblay, J., McKay, D., Yamaguchi, Y., Leung, K., Balanon, A., ... & Birchfield, S. (2018). NDDS： NVIDIA deep learning dataset synthesizer.
35. Indoor CVPR 2009 Image Dataset, http：//web.mit.edu/torralba/www/indoor.html
36. Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., Leach, P., & Bern-ers-Lee, T. (1999). Hypertext transfer protocol--HTTP/1.1 (No. rfc2616).
37. Josefsson, S. (2006). The base16, base32, and base64 data encodings (No. rfc4648).
38. Hyndman, R. J. (2011). Moving Averages.
39. Dantzig, G. B. (2002). Linear programming. Operations research, 50(1), 42-47.
40. 楊敏生 (1994)，模糊理論簡介，數學傳播中央研究院數學研究所，台北。
41. Stehman, S. V. (1997). Selecting and interpreting measures of thematic classifica-tion accuracy. Remote sensing of Environment, 62(1), 77-89.
42. 潘南飛、黃冠智，模糊線性規劃用於專案排成分析之研究，高雄應用科技大學土木工程與防災科技研究所，工程科技與教育學刊。
43. Sim, I. (2019). Mobile devices and health. New England Journal of Medi-cine, 381(10), 956-968.
44. Nickolls, J., & Dally, W. J. (2010). The GPU computing era. IEEE micro, 30(2), 56-69.
45. Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016). Edge computing： Vision and challenges. IEEE internet of things journal, 3(5), 637-646.
46. Popovski, P., Trillingsgaard, K. F., Simeone, O., & Durisi, G. (2018). 5G wireless network slicing for eMBB, URLLC, and mMTC： A communication-theoretic view. Ieee Access, 6, 55765-55779.