臺灣博碩士論文加值系統

近年來工業4.0 發展快速，將各領域技術結合並運用，提升智慧製造之能力，為此各國都因應工業4.0 提出相對應之發展政策。台灣針對工業4.0 提出「生產力 4.0」之發想，使台灣各產業針對智慧製造之技術快速發展，提升台灣製造產業於國際之競爭力。

因應智慧製造技術的需求提升，機械手臂被廣泛應用其中，主要因機械手臂其生產能力穩定，可為產線帶來眾多效益。機械手臂若需配合智慧工廠之多種情境變化，可透過人工進行教調達成目的，但若可使機械手臂智慧化，可應因不同環境或不同製造程序自行調整而不需頻繁透過人力介入，就可大幅降低人工成本且提生產線之製造效率。

本研究嘗試整合機械手臂與相機配合產線所常用之輸送帶，搭配視覺辨識演算法，為使機械手臂智慧化，可夾取於輸送帶上不同位置物件，並針對運行效率以及準確度，對多種演算法模型以及移動路徑進行測試紀錄並將其結果進行分析。

本研究利用機械手臂結合視覺辨識演算法，搭配相機及輸送帶，建立一智慧化之物件取放系統，可針對產線之彈性位置物件進行夾取，優化機械手臂之取放能力，利用此系統針對多種演算法模型及路徑移動方式進行分析並得出各者優異，並針對多種情境進行建議，達到多種情境之需求及最佳成果。

In recent years, industry 4.0 has developed rapidly, combining and applying technologies in various fields to enhance the ability of smart manufacturing. Therefore, many countries have put forward corresponding development policies according to industry 4.0. Taiwan has put forward the idea of "productivity 4.0" for industry 4.0, so that Taiwan's various industries can develop the technology of smart manufacturing rapidly and enhance the competitiveness of Taiwan's manufacturing industry in the international.

In response to the increasing demand of smart manufacturing technology, robotic arm is widely used, mainly because of its stable production capacity, which can bring many benefits to the production line. If the robotic arm needs to meet the various situation changes of the smart factory, it can be taught and adjusted manually to achieve the purpose. However, if the robotic arm arm can be made intelligent, it can be adjusted according to different environments or different manufacturing procedures without frequent human intervention, which can greatly reduce the labor cost and improve the manufacturing efficiency of the production line.

This study integrates the robotic arm and camera to cooperate with the conveyor belt commonly used in the production line, and is equipped with a visual recognition algorithm. In order to make the robotic arm intelligent, it can clamp objects at different positions on the conveyor belt. The algorithm model and the movement path are tested and recorded and the results are analyzed.

This study uses a robotic arm combined with a visual recognition algorithm, with cameras and conveyor belts, to establish an intelligent object pick-and-place system, which can pick up objects at elastic positions in the production line and optimize the pick-and-place operation of the robotic arm. Using this system analyzed various algorithm models and the movement path and concluded that each is excellent, and made recommendations for various scenarios to achieve the needs and best results of various scenarios.

中文摘要 ...... i
英文摘要 ...... ii
誌謝 ...... iv
目錄 ...... v
表目錄 ...... vii
圖目錄 ...... viii
第1章 緒論 ...... 1
1.1 研究背景 ...... 1
1.2 研究動機與目的 ...... 1
1.3 研究流程 ...... 2
第2章 文獻探討 ...... 3
2.1 深度學習 ...... 3
2.2 YOLO演算法 ...... 5
2.2.1 YOLOv3 ...... 7
2.2.2 YOLOv4 ...... 9
2.3 機械手臂 ...... 13
第3章 研究方法 ...... 15
3.1 研究架構 ...... 15
3.2 機械手臂控制 ...... 17
3.3 YOLO模型訓練 ...... 18
3.3.1 訓練資料 ...... 18
3.3.2 訓練參數與模型結果 ...... 19
第4章 研究分析與成果 ...... 24
4.1 研究設備 ...... 24
4.1.1 運算設備 ...... 24
4.1.2 UR5協作機械手臂 ...... 24
4.1.3 輸送帶控制模組 ...... 24
4.1.4 相機 ...... 25
4.2 二維路徑夾取研究分析 ...... 25
4.2.1 二維路徑演算法 ...... 26
4.2.2 二維路徑運作時間效率分析 ...... 27
4.2.3 二維路徑準確度分析 ...... 29
4.2.4 二維路徑及模型情境建議 ...... 31
4.3 三維路徑夾取研究分析 ...... 32
4.3.1 三維路徑演算法 ...... 32
4.3.2 三維路徑運作時間效率分析 ...... 32
4.3.3 三維路徑準確度分析 ...... 32
4.3.4 三維路徑情境建議 ...... 33
4.4 多加工站點移動方式時間效率分析 ...... 33
第5章 結論與未來發展 ...... 35
5.1 結論 ...... 35
5.2 未來展望 ...... 35
參考文獻 ...... 37
附錄一 ...... 39
附錄二 ...... 50
附錄三 ...... 61
附錄四 ...... 63
Extended Abstract ...... 66

[1]Industrie 4.0 Working Group, Recommendations for implementing the strategic initiative INDUSTRIE 4.0, Final report, April 2013.
[2]楊錦洲. (2020). 工業 4.0 與智能製造. 品質月刊, 56(7), 6-9.
[3]LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. nature, 521(7553), 436-444.
[4]江瑞璋, & 孫沛立. (2021). 以連續角度光源及卷積神經網路檢測光影變化箔膜影像. 中華印刷科技年報, 42-49.
[5]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788).
[6]Redmon, J., & Farhadi, A. (2018). Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767.
[7]Lin, Tsung-Yi, et al. "Feature pyramid networks for object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[8]Redmon,J.,&Farhadi,A.(2017). YOLO9000: better, faster, stronger. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7263-7271).
[9]Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. M. (2020). Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.
[10]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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops (pp. 390-391).
[11]Liu, S., Qi, L., Qin, H., Shi, J., & Jia, J. (2018). Path aggregation network for instance segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 8759-8768).
[12]Javaid, M., Haleem, A., Singh, R. P., & Suman, R. (2021). Substantial Capabilities of Robotics in Enhancing Industry 4.0 implementation. Cognitive Robotics.
[13]IFR World Robotics 2021 reports,October 2021.
[14]Andronas, D., Argyrou, A., Fourtakas, K., Paraskevopoulos, P., & Makris, S. (2020). Design of Human Robot Collaboration workstations–Two automotive case studies. Procedia Manufacturing, 52, 283-288.
[15]Bartoš, M., Bulej, V., Bohušík, M., Stanček, J., Ivanov, V., & Macek, P. (2021). An overview of robot applications in automotive industry. Transportation Research Procedia, 55, 837-844.
[16]Zhang, B., Xie, Y., Zhou, J., Wang, K., & Zhang, Z. (2020). State-of-the-art robotic grippers, grasping and control strategies, as well as their applications in agricultural robots: A review. Computers and Electronics in Agriculture, 177, 105694.
[17]梁金銅(2017).機械手臂輔助手術在台灣施行的現況.台灣醫學,21(2) ,168-172.