跳到主要內容

臺灣博碩士論文加值系統

(44.200.194.255) 您好!臺灣時間:2024/07/24 05:14
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:黃丞陽
研究生(外文):Cheng-Yang Huang
論文名稱:使用YOLO深度學習網路於滾珠螺桿工程圖辨識與其關鍵參數提取之研究
論文名稱(外文):A Study on The Recognition and Key Parameters Extraction for The Engineering Drawings of Ball Screw Using YOLO Deep Learning Network
指導教授:吳俊霖吳俊霖引用關係
指導教授(外文):Jiunn-Lin Wu
口試委員:林詠章劉志俊
口試日期:2024-01-31
學位類別:碩士
校院名稱:國立中興大學
系所名稱:資訊工程學系所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2024
畢業學年度:112
語文別:中文
論文頁數:41
中文關鍵詞:深度學習物件偵測YOLO光學字元辨識滾珠螺桿洪水填充演算法
外文關鍵詞:Deep learningObject detectionYOLOTessercat OCRBall ScrewFlood Fill Algorithm
相關次數:
  • 被引用被引用:0
  • 點閱點閱:147
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
滾珠螺桿是關鍵的機械組件,它是由滾桿、螺帽、鋼珠所組成,鋼珠置於螺桿和螺帽之間運作,有效地將旋轉運動轉換為直線運動,具有高壽命、低污染、低摩擦、高定位精度等優點,廣泛應用在精密機械和自動化設備等。滾珠螺桿可以細分四個主要部件,分別為法蘭、螺帽、牙部、肩部。
工程圖是以線條與符號所呈現的一種技術檔,確保產品的性能以及尺寸有達到需求,數學標示數量關係,文字註解表示各種說明,用圖形表示設計結構,而圖面中的標註可再細分參數與標註線。
傳統通過人員來解讀滾珠螺桿工程圖,圖面上的線條與文字數量繁多,確認參數位置並辨識出其數值,以人工方式相當均耗時。解讀滾珠螺桿工程圖,除了要有圖學的知識還要有滾珠螺桿的背景知識,才能在滾珠螺桿工程圖上識別出部件的位置,從而找到關鍵參數,分別為螺桿總長、螺帽外徑、法蘭外徑、右側肩部數量、左側肩部數量。
人員的專業性不足又或長時間工作導致精神疲勞等原因,都可能出現解讀滾珠螺桿圖面時出現錯誤,而不管在銷售端還是製程端,其中解讀出來的關鍵參數的若是錯誤,需要耗費額外的時間、金錢來處理,對於公司有各方面的負面影響。
因此,本研究提出基於YOLOv7深度學習的滾珠螺桿辨識系統,旨在於加速滾珠螺桿圖面關鍵參數的識別速度以及減少人為錯誤的可能性,進而協助人員進行後續的工作作業。由於實際圖面有機密性問題,因此從開放的官方網站,使用不同的外徑、導程、螺帽型式等條件進行資料蒐集。接下來,依訓練及應用需求將資料轉化為影像檔並對資料註記,完成深度學習模型訓練。辨識系統有四個階段,來達成滾珠螺桿工程圖關鍵參數的提取。由實驗結果顯示,YOLOv7偵測滾珠螺桿部件(大圖形)以及參數的辨識率達99%。特別在具有相似圖形與一致標註習慣的滾珠螺桿工程圖時,關鍵參數辨識率以及整體辨識率均為100%。
Ball screws are key mechanical components composed of a screw, nut, and balls that operate between the screw and nut, effectively converting rotational movement into linear motion. They offer advantages such as high durability, low pollution, low friction, and high positioning accuracy, making them widely used in precision machinery and automation equipment. Ball screws can be divided into four parts: flange, nut, thread, and shoulder.
An engineering drawing, represented by lines and symbols, ensures product performance and dimensional requirements are met. It uses mathematical notation to express quantitative relationships and textual annotations for various explanations, representing design structures graphically, with details further divided into parameters and annotation lines.
Traditionally, interpreting ball screw engineering drawings, with their numerous lines and text, required confirming parameter locations and identifying their values manually, a time-consuming process. Besides having knowledge of drafting, background knowledge of ball screws is also necessary to identify component locations on the drawings and find key parameters such as total screw length, nut outer diameter, flange outer diameter, the number of right and left shoulders.
Lack of expertise or mental fatigue from long hours can lead to errors in interpreting ball screw drawings. Mistakes in key parameters, whether in sales or manufacturing, can lead to additional time and cost, negatively impacting the company.
Therefore, this study introduces a ball screw recognition system based on YOLOv7 deep learning, aimed at accelerating the identification speed of key parameters in ball screw drawings and reducing the possibility of human error, thereby assisting personnel in subsequent tasks. Due to confidentiality issues with actual drawings, data was collected from official websites using different conditions such as outer diameter, lead, and nut type. The data was then converted into image files and annotated for deep learning model training. The recognition system comprises four stages to extract key parameters from ball screw engineering drawings. Experimental results show YOLOv7's detection of ball screw components and parameters achieves 99% accuracy. For ball screw drawings with similar graphics and consistent labels, key parameters and overall accuracy reach 100%.
第一章 緒論 1
1.1 研究背景及動機 1
1.2 論文架構 5
第二章 文獻探討 6
2.1 基於深度學習的2D工程圖圖形辨識方式(Engineering Drawing Graphic Recognition Method) 6
2.2 提取參數標註方法 (Size Annotation Method) 7
2.2.1 基於聚類演算法(Based on Clustering Algorithm) 7
2.2.2 基於深度學習方法 (Deep Learning Method) 8
第三章 研究方法 10
3.1 問題陳述與研究目標 10
3.2 方法流程圖 10
3.3 資料預處理方法 11
3.4 滾珠螺桿關鍵參數的辨識系統 12
3.4.1 部件定位 13
3.4.2 YOLO 物件偵測模型 13
3.4.3 YOLOv7 14
3.4.4 滾珠螺桿工程圖大圖形的差異處 15
3.4.5 線條精簡 18
3.4.6 提取參數的困難點 19
3.4.7 辨識參數的方式 20
3.4.8 提取關鍵參數的判斷 22
第四章 實驗結果與討論 24
4.1 實驗環境 24
4.2 YOLOv7 超參數設定 24
4.3 滾珠螺桿資料集 25
4.3.1 部件資料集 25
4.3.2 參數資料集 26
4.3.3 滾珠螺桿工程圖的註記 26
4.4 滾珠螺桿資料準的細節 29
4.5 模型評估指標 30
4.6 模型性能評估與分析 31
4.7 模型識別的效果 33
4.7.1 部件模型的效果 33
4.7.2 參數模型的效果 34
4.8 關鍵參數評估指標 35
4.9 關鍵參數辨識評估 35
4.10 實驗結果可視化 36
第五章 結論與未來展望 39
參考文獻 40
[1]莊瀅芯, "滾珠螺桿產業現況與發展趨勢," (in 繁體中文), 機械工業雜誌, no. 430, pp. 40-45, 2019. [Online]. Available: Article/Detail?docID=P20171221002-201901-201901020022-201901020022-40-45.
[2]HIWIN Technologies Corp., "滾珠螺桿-圖面下載," HIWIN, https://www.hiwin.tw/cadpreview/ballscrew.aspx, Nov. 2023 (accessed Nov. 16, 2023).
[3]Y. Yin, Z. Meng, and S. Li, "Feature extraction and image recognition for the electrical symbols based on Zernike moment," in 2017 IEEE 2nd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), 25-26 March 2017 2017, pp. 1031-1035, doi: 10.1109/IAEAC.2017.8054170.
[4]T. Cheng, J. Khan, H. Liu, and D. Y. Y. Yun, "A symbol recognition system," in Proceedings of 2nd International Conference on Document Analysis and Recognition (ICDAR '93), 20-22 Oct. 1993 1993, pp. 918-921, doi: 10.1109/ICDAR.1993.395587.
[5]Y. Luo, T. Yu, J. Zheng, and Y. Ding, "Design of engineering drawing recognition system based on Yolo V4," in 2022 IEEE 6th Information Technology and Mechatronics Engineering Conference (ITOEC), 4-6 March 2022 2022, vol. 6, pp. 1221-1225, doi: 10.1109/ITOEC53115.2022.9734453.
[6]S. Zeng, Y. Luo, and Y. Ding, "Large-scale Network Survivability Association Model based on Set Pair Analysis Theory," presented at the 2022 IEEE 6th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC ), 2022.
[7]Y. Liu, Y. Wang, Z. Liu, Y. Liu, and X. Shen, "Research on Improved YOLOv5-Based Small Component Recognition Method of Electrical Drawings," in 2023 8th International Conference on Signal and Image Processing (ICSIP), 8-10 July 2023 2023, pp. 64-68, doi: 10.1109/ICSIP57908.2023.10270882.
[8]A. Mane, R. Adhikari, S. Gadgil, and N. Raykar, "Investigating Application of Machine Learning in Identification of Polygon Shapes for Recognition of Mechanical Engineering Drawings," in 2019 International Conference on Nascent Technologies in Engineering (ICNTE), 4-5 Jan. 2019 2019, pp. 1-6, doi: 10.1109/ICNTE44896.2019.8945917.
[9]T. P. Hong, H. W. Chiu, S. F. Huang, and Y. T. Chen, "Deep-learning-based Extraction of Electronic Component Parameters from Datasheets," in 2021 IEEE International Conference on Big Data (Big Data), 15-18 Dec. 2021 2021, pp. 5501-5506, doi: 10.1109/BigData52589.2021.9672035.
[10]B. Scheibel, J. Mangler, and S. Rinderle-Ma, "Extraction of dimension requirements from engineering drawings for supporting quality control in production processes," Computers in Industry, vol. 129, Aug 2021, Art no. 103442, doi: 10.1016/j.compind.2021.103442.
[11]黃奕群, "2D工程圖之深度學習的整合應用," 碩士, 機械工程學系, 中原大學, 桃園縣, 2022. [Online]. Available: https://hdl.handle.net/11296/ngew4b
[12]A. Sinha, J. Bayer, and S. S. Bukhari, "Table Localization and Field Value Extraction in Piping and Instrumentation Diagram Images," in 2019 International Conference on Document Analysis and Recognition Workshops (ICDARW), 22-25 Sept. 2019 2019, vol. 1, pp. 26-31, doi: 10.1109/ICDARW.2019.00010.
[13]B. Seraogi et al., "Automatic Orientation Correction of AEC Drawing Documents," in 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), 9-15 Nov. 2017 2017, vol. 02, pp. 9-10, doi: 10.1109/ICDAR.2017.252.
[14]W. Khallouli, R. Pamie-George, S. Kovacic, A. Sousa-Poza, M. Canan, and J. Li, "Leveraging Transfer Learning and GAN Models for OCR from Engineering Documents," in 2022 IEEE World AI IoT Congress (AIIoT), 6-9 June 2022 2022, pp. 015-021, doi: 10.1109/AIIoT54504.2022.9817319.
[15]C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, "YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 7464-7475.
連結至畢業學校之論文網頁點我開啟連結
註: 此連結為研究生畢業學校所提供,不一定有電子全文可供下載,若連結有誤,請點選上方之〝勘誤回報〞功能,我們會盡快修正,謝謝!
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊