職業傷害的問題已經日益受到重視，專家學者為此問題提供了像是肌電圖檢查、超音波、各式評估量表、問卷調查等解決方式，其中以便利性而言，業界大多採用國際通用量表來進行職業傷害的評估，如勞工姿勢評估系統(OWAS）、快速上肢評估法（RULA）、快速全身肢體評估法（REBA）等量表，各量表會依不同工作環境上做使用，然而在評估上會因主觀判斷而產生誤差，且需要耗費相當的人力以及時間。影像式動作擷取技術是近年來常用於分析人體動作的技術，此技術是利用拍攝到的RGB影像進行影像分析，進而取得人體關鍵點的資訊，並建構出人體骨架模型。
為了滿足不同工作環境上的需求，以及解決評估時間和成本上的浪費，本研究欲建立一套自動化量表選擇即時評估系統，將選用業界最常用的REBA、OWAS、RULA量表來進行評估，透過網路資源收尋：維修、搬運、組裝、清潔、電腦操作、開車等6種作業，共15支影片。經由影像式截取技術Openpose進行處理，輸出關鍵點座標，並搭配向量的方式來計算出人體關節角度。其中，在關節角度上，以即時呈現的方式。藉此判斷此作業影片是否有發生屏蔽的干擾，並透過三次樣條插值的方式來降噪。在量表選擇上皆符合勞研所的規範，利用多個關節角度來進行動作特性判別，以選擇出合適的量表，進而得到更適切的職業傷害風險評估，最終輸出量表分級結果，並回傳給使用者在此工作姿勢下的風險程度。
自動化量表選擇評估系統對應於不同工作環境上，選擇了更適合的量表進行評估，相較於過往專家，在選擇結果上更能滿足各量表的設計理念以及規範；在量表評估上，因網路取材的關係，將不考慮施力、負載重量兩項評估因子，在分析結果中，節省了專家挑選評估圖片以及評估的時間。並預期透過即時呈現的方式，達到長期監控與預防職業傷害的發生。

The issue of occupational injuries has been increasingly valued. Experts and scholars have provided solutions to this problem, such as electromyography, ultrasound, various evaluation scales, questionnaires, etc. Among them, in terms of convenience, most of the industry adopts general international scales that are used to evaluate occupational injuries, such as REBA, OWAS, RULA, and other scales. Each scale is used according to different working environments. However, the evaluation causes errors due to subjective judgment and is quite labor intensive and time-consuming. Image-based motion capture technology is a technology commonly used in recent years to analyze human body motions. This technology uses the captured RGB images for image analysis to obtain information about key points of the human body and construct a human skeleton model.
To meet the needs of different working environments and solve the waste of evaluation time and cost. This study aims to build an automated scale selection real-time evaluation system. The REBA, OWAS, RULA scales are most commonly used for evaluation in the industry. Find six kinds of working conditions (maintenance, handling, assembly, cleaning, computer user, driving) of fifteen videos on the internet. Through the image-based interception technology Openpose for processing, output the coordinates of key points, and match the vector to calculate the angle of the human joint. Among them, the joint angle is presented in real time. It is used to determine whether there is interference in the operation video, and the noise is reduced by cubic spline interpolation. The selection results are consistent with the specifications of the previous literature. Use multiple joint angles to determine motion characteristics to select a suitable scale, and then get a more reasonable occupational injury risk assessment. Finally, the scale results are output and returned to the user.
The automatic scale selection and evaluation system correspond to different working environments. It selects a more suitable scale for evaluation, reducing the time for experts to select and evaluate. Compared with past experts, the selection results can better meet the design concepts and specifications of each scale. In terms of scale evaluation, due to the relationship of the network, the two evaluation factors of force and load weight will not be considered. In the analysis result, we saved the time for experts to select and evaluate pictures in 15 assignment videos. It achieves long-term monitoring and prevention of occupational injuries through the real-time presentation.

摘要 II
目錄 V
表目錄 VII
圖目錄 VIII
第一章 緒論 1
1.1. 研究背景與動機 1
1.2. 研究目的與範圍 5
1.3. 研究架構與流程 6
第二章 文獻回顧 7
2.1. 工作環境與職業傷害的關係 7
2.2. 影像式截取技術以及Openpose的選擇 9
2.2.1. Openpose 13
2.3. 動作特性判別特徵選擇 16
2.4. 量表選擇與介紹 17
2.4.1. 勞工姿勢評估系統(OWAS) 18
2.4.2. 快速上肢評估法(RULA) 20
2.4.3. 快速全身肢體評估法(REBA) 22
2.5. 小結 25
第三章 研究方法 26
3.1. 問題定義與描述 26
3.2. 系統架構與資料獲取 26
3.2.1. 資料獲取 28
3.2.2. 關節角度計算 34
3.2.3. 量表選擇 36
3.2.4. 量表評估系統 38
3.3. 實驗流程 38
第四章 實驗結果 39
4.1. 各項作業角度評估結果 39
4.1.1. 可視化關節角度 40
4.1.2. 受遮蔽關節角度訊號處理 42
4.2. 量表選擇結果 43
4.3. 量表評估結果 45
第五章 討論 50
5.1. 關節角度可行性與正確性 50
5.2. 自動化量表評估系統之成效 51
5.3. 研究限制 54
第六章 結論與未來方向 55
附錄(一) 56
參考文獻 104

中文部分：
1. 職業安全衛生法。
取自：https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=N0060001
2. 勞動部職業安全衛生署，107年全國職業傷病診斷治網路職業疾病通報數統計概況。
取自：https://www.osha.gov.tw/1106/1113/1114/24256/
3. 勞動環境安全衛生狀況認知調查-2016年 ILOSH105-A309。
取自：https://www.books.com.tw/products/0010803210
4. 勞動部職業案全衛生署(2014)，人因性危害預防計畫指引。
取自：https://www.osha.gov.tw/media/4637/%e4%ba%ba%e5%9b%a0%e6%80%a7%e5%8d%b1%e5%ae%b3%e9%a0%90%e9%98%b2%e8%a8%88%e7%95%ab%e6%8c%87%e5%bc%95.doc
5. 勞動部勞動及職業安全衛生研究所(2014)，「人因工程肌肉骨骼傷病預防指引」。
取自: http://si.secda.info/stchang/wp-content/uploads/2014/10/0.pdf
6. 動作捕捉的分類與發展。
取自：https://kknews.cc/zh-tw/tech/29mr99.html
外文部分：
1. Ansari, N., & Sheikh, M. (2014). Evaluation of work Posture by RULA and REBA: A Case Study. IOSR Journal of Mechanical and Civil Engineering, 11(4), 18-23.
2. Cao, Z., Hidalgo, G., Simon, T., Wei, S.-E., & Sheikh, Y. (2018). OpenPose: realtime multi-person 2D pose estimation using Part Affinity Fields. arXiv preprint arXiv:1812.08008.
3. Chen, Y., Wang, Z., Peng, Y., Zhang, Z., Yu, G., & Sun, J. (2018). Cascaded pyramid network for multi-person pose estimation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,
4. Coenen, P., Kingma, I., Boot, C. R., Faber, G. S., Xu, X., Bongers, P. M., & Van Dieen, J. H. (2011). Estimation of low back moments from video analysis: A validation study. Journal of biomechanics, 44(13), 2369-2375.
5. Das, B. (2014). Prevalence of work-related musculoskeletal disorders among the brick field workers of West Bengal, India. Archives of environmental & occupational health, 69(4), 231-240.
6. Das, B., Ghosh, T., & Gangopadhyay, S. (2012). Assessment of ergonomic and occupational health-related problems among female prawn seed collectors of Sunderbans, West Bengal, India. International Journal of Occupational Safety and Ergonomics, 18(4), 531-540.
7. Etemadinezhad, S., Ranjbar, F., & Gorji, M. (2013). Posture Analysis by OWAS Method and Prevalence of Musculoskeletal Disorders using Nordic Questionnaire among Workers of Sourak Tobacco Factory in 2013. Iranian journal of health sciences, 1(2), 89-94.
8. Evangelidis, G., Singh, G., & Horaud, R. (2014). Skeletal quads: Human action recognition using joint quadruples. 2014 22nd International Conference on Pattern Recognition,
9. Fani, H., Mirlohi, A., Hosseini, H., & Herperst, R. (2018). Swim stroke analytic: Front crawl pulling pose classification. 2018 25th IEEE International Conference on Image Processing (ICIP),
10. Ghazal, S., & Khan, U. S. (2018). Human posture classification using skeleton information. 2018 International Conference on Computing, Mathematics and Engineering Technologies (iCoMET),
11. Go, D. S., Hyung, H. J., Lee, D. W., & Yoon, H. U. (2018). Andorid Robot Motion Generation Based on Video-Recorded Human Demonstrations. 2018 27th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN),
12. Insafutdinov, E., Pishchulin, L., Andres, B., Andriluka, M., & Schiele, B. (2016). Deepercut: A deeper, stronger, and faster multi-person pose estimation model. European Conference on Computer Vision,
13. Krüger, J., & Nguyen, T. D. (2015). Automated vision-based live ergonomics analysis in assembly operations. CIRP Annals, 64(1), 9-12.
14. Lee, T.-H., & Han, C.-S. (2013). Analysis of working postures at a construction site using the OWAS method. International Journal of Occupational Safety and Ergonomics, 19(2), 245-250.
15. Li, W., Zhang, Z., & Liu, Z. (2010). Action recognition based on a bag of 3d points. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition-Workshops,
16. Malaisé, A., Maurice, P., Colas, F., & Ivaldi, S. (2019). Activity recognition for ergonomics assessment of industrial tasks with automatic feature selection. IEEE Robotics and Automation Letters, 4(2), 1132-1139.
17. Mathie, M. (2003). Monitoring and interpreting human movement patterns using a triaxial accelerometer. University of New South Wales Sydney.
18. McKinley, S., & Levine, M. (1998). Cubic spline interpolation. College of the Redwoods, 45(1), 1049-1060.
19. Moussavi-Najarkola, S. A., & Mirzaei, R. (2012). Assessment of musculoskeletal loads of electric factory workers by rapid entire body assessment. Health Scope, 1(2), 71-79.
20. Norkin, C. C., & White, D. J. (2016). Measurement of joint motion: a guide to goniometry. FA Davis.
21. Ohn-Bar, E., & Trivedi, M. (2013). Joint angles similarities and HOG2 for action recognition. Proceedings of the IEEE conference on computer vision and pattern recognition workshops,
22. Padilla, B. E. O., Glushkova, A., Menychtas, D., & Manitsaris, S. (2019). Designing a web-based automatic ergonomic assessment using motion data. Proceedings of the 12th ACM International Conference on PErvasive Technologies Related to Assistive Environments,
23. Pal, A., & Dhara, P. C. (2017). Evaluation of work-related musculoskeletal disorders and postural stress of female “Jari” workers. Indian journal of occupational and environmental medicine, 21(3), 132.
24. Pal, A., & Dhara, P. C. (2018). Work related musculoskeletal disorders and postural stress of the women cultivators engaged in uprooting job of rice cultivation. Indian journal of occupational and environmental medicine, 22(3), 163.
25. Qu, Y., Hwang, J., Lee, K.-S., & Jung, M.-C. (2012). The effect of camera location on observation-based posture estimation. Ergonomics, 55(8), 885-897.
26. Rafie, F., Zamani Jam, A., Shahravan, A., Raoof, M., & Eskandarizadeh, A. (2015). Prevalence of upper extremity musculoskeletal disorders in dentists: symptoms and risk factors. Journal of environmental and public health, 2015.
27. Tinubu, B. M., Mbada, C. E., Oyeyemi, A. L., & Fabunmi, A. A. (2010). Work-related musculoskeletal disorders among nurses in Ibadan, South-west Nigeria: a cross-sectional survey. BMC Musculoskeletal disorders, 11(1), 12.
28. Vemulapalli, R., Arrate, F., & Chellappa, R. (2014). Human action recognition by representing 3d skeletons as points in a lie group. Proceedings of the IEEE conference on computer vision and pattern recognition,
29. Wang, H. L., & Lee, Y.-J. (2019). Occupational evaluation with Rapid Entire Body Assessment (REBA) via imaging processing in field. Human Factors Society Conference,
30. Zhang, Z. (2000). A flexible new technique for camera calibration. IEEE Transactions on pattern analysis and machine intelligence, 22.