臺灣博碩士論文加值系統

近年來，人工智慧領域發展迅速，為許多產業帶來了各種應用，能夠有效減輕相關成本，並提升效率與精準度。體育產業結合人工智慧科技將是未來趨勢，人工智慧技術可以用於分析運動員的訓練資料，提供個性化的訓練方案；在比賽中，人工智慧技術可以用於裁判輔助，提高比賽的公正性和準確性。這些應用不僅能夠提高運動表現，還能夠為體育產業帶來更多的商機和發展機遇。根據教育部體育署統計，在台灣民眾最常從事的運動類型中，羽球是球類運動中最多人參與的，這表明羽球在台灣的受歡迎程度，也為羽球在運動科技發展中的應用提供了廣大的市場和需求。本研究設計個人即可進行的羽球對打訓練系統，並且能夠蒐集選手相關資料，分析並評估選手的訓練效益。具體來說，本研究設計之系統會使用相機辨識出羽球場地範圍，並且劃分成九個區域後用於判斷選手即時位置，接著系統可以透過控制發球機，決定發球點。為了增加趣味性，本研究將設計出不同難度時的發球點選擇方法，選手可以依據自身狀況選擇合適的方法訓練。而選手資料收集的部分，本研究將透過姿態估計演算法來計算選手的身體節點，並計算出選手的動作變化，這些資料除了可以用於設計更靈活的發球點選擇方法外，也能回饋給選手，判斷姿勢是否正確等。最後，本論文評估了所使用之神經網路演算法的效能，並且使用世界羽球聯盟之比賽影片進行系統模擬，再透過實際場域佈署進行流程測試，透過記錄不同難易度下選手的反應狀況，計算訓練多次後對於羽球技能是否有精進，輔以姿態估計分析出選手動作是否有改善空間，以證系統有提供一定程度的效益。

In recent years, the rapid development of artificial intelligence (AI) has led to various applications across numerous industries, reducing costs and enhancing efficiency and accuracy. Integrating AI into the sports industry is a future trend, as it can analyze athletes' training data to provide personalized training programs and assist referees to improve fairness and accuracy in competitions. These applications enhance athletic performance and create new business opportunities for the sports industry. According to the Sports Administration of the Ministry of Education in Taiwan, badminton is the most popular ball sport among the public. This popularity indicates a broad market and demand for sports technology applications in badminton.
This research aims to design an individual-use badminton training system. The system collects player data, analyzes it, and evaluates training effectiveness. It identifies badminton court boundaries using a camera, divides the court into nine zones for real-time player positioning, and controls the shuttlecock machine to determine the serve point. To increase training enjoyment, different difficulty levels for serve points are designed, allowing players to choose the appropriate method based on their skill level. Player data is collected using a pose estimation algorithm to calculate body keypoints and analyze movements. This data helps design flexible serve point methods and provides feedback on posture correctness. Finally, the paper evaluates the performance of the neural network algorithms used and simulates the system with Badminton World Federation (BWF) match videos. The system is deployed for real-world testing, recording player responses under different difficulty levels and analyzing skill improvement after training sessions. Pose estimation identifies areas for improvement, demonstrating that the system provides significant benefits.

摘要...i
Abstract...ii
誌謝...iii
目錄...iv
表目錄...vi
圖目錄...vii
第一章 緒論...1
1.1 研究動機與目的...1
1.2 相關研究...2
1.3 章節安排...4
第二章 文獻回顧...5
2.1 影像處理...6
2.1.1 邊緣檢測...6
2.1.2 霍夫變換...7
2.1.3 Point in Polygon...9
2.1.4 單應性矩陣...10
2.2 機器學習和深度學習...11
2.2.1 神經網路...12
2.2.2 卷積神經網路...15
2.2.3 遞迴神經網路...16
2.2.4 長短期記憶...16
2.3 You Only Look Once(YOLO)...17
2.3.1 YOLOv9...19
2.4 姿態估計...20
2.4.1 MediaPipe...21
2.4.2 OpenPose...23
2.5 物聯網...25
第三章 系統設計...27
3.1 系統架構...27
3.1.1 硬體架構...27
3.1.2 軟體架構...28
3.2 系統流程...29
3.2.1 啟動系統...30
3.2.2 APP端連接系統...30
3.2.3 APP其他頁面流程...31
3.3 實驗方法...33
3.3.1 場地辨識...33
3.3.2 場地分割...33
3.3.3 定位球員演算法...38
3.3.4 難易度設計...38
3.3.5 姿態估計分析...40
第四章 實驗與結果...42
4.1 場地辨識與分割結果...42
4.2 模型訓練結果...43
4.2.1 模型評估方法...43
4.2.2 K-Fold交叉驗證...45
4.2.3 YOLO...46
4.2.4 LSTM...49
4.3 演算法模擬...53
4.4 軟體介面...55
4.4.1 首頁...56
4.4.2 訓練頁面...58
4.4.3 歷史紀錄頁面...59
4.4.4 姿態分析頁面...60
4.5 實驗結果...62
4.5.1 硬體佈署...62
4.5.2 測試結果...62
第五章 結論與未來展望...64
參考文獻...65
Extended Abstract...68

[1]教育部體育署. (2023). i運動資訊平台-運動現況統計. Retrieved from https://isports.sa.gov.tw/Apps/Download.aspx?SYS=TIS&MENU_CD=M07&ITEM_CD=T01&MENU_PRG_CD=4&ITEM_PRG_CD=2
[2]Wang, Y., Zhao, Y., Chan, R. H. M., & Li, W. J. (2018). Volleyball skill assessment using a single wearable micro inertial measurement unit at wrist. IEEE Access, 6, 13758-13765. https://doi.org/10.1109/ACCESS.2018.2792220
[3]Pan, T. -Y., Tsai, W. -L., Chang, C. -Y., Yeh, C. -W., & Hu, M. -C. (2022). A hierarchical hand gesture recognition framework for sports referee training-based EMG and accelerometer sensors. IEEE Transactions on Cybernetics, 52(5), 3172-3183. https://doi.org/10.1109/TCYB.2020.3007173
[4]Strohrmann, C., Harms, H., Kappeler-Setz, C., & Troster, G. (2012). Monitoring kinematic changes with fatigue in running using body-worn sensors. IEEE Transactions on Information Technology in Biomedicine, 16(5), 983-990. https://doi.org/10.1109/TITB.2012.2201950
[5]Ma, H., & Ding, X. (2022). Robust automatic camera calibration in badminton court recognition. In 2022 IEEE Asia-Pacific Conference on Image Processing, Electronics and Computers (IPEC) (pp. 893-898). https://doi.org/10.1109/IPEC54454.2022.9777532
[6]De Alwis, A. P. G., Dehikumbura, C., Konthawardana, M., Lalitharatne, T. D., & Dassanayake, V. P. C. (2020). Design and development of a badminton shuttlecock feeding machine to reproduce actual badminton shots. In 2020 5th International Conference on Control and Robotics Engineering (ICCRE) (pp. 73-77).
[7]Chen, Y. -T., Yang, J. -F., & Tu, K. -C. (2021). Smart badminton detection system based on Scaled-YOLOv4. In 2021 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) (pp. 1-2). https://doi.org/10.1109/ISPACS51563.2021.9651036
[8]Huang, T., Li, Y., & Zhu, W. (2021). An auxiliary training method for single-player badminton. In 2021 16th International Conference on Computer Science & Education (ICCSE) (pp. 441-446). https://doi.org/10.1109/ICCSE51940.2021.9569592
[9]Raina, A., Lakshmi, T. G., & Murthy, S. (2017). CoMBaT: Wearable technology based training system for novice badminton players. In 2017 IEEE 17th International Conference on Advanced Learning Technologies (ICALT) (pp. 153-157). https://doi.org/10.1109/ICALT.2017.96
[10]Chiu, C. -H., Su, J. -L., & Lin, C. -M. (2023). The prediction of badminton flight trajectory based on an intelligent compensator. IEEE Access, 11, 32261-32271. https://doi.org/10.1109/ACCESS.2023.3262554
[11]Tai, W. -S., & Liu, K. -H. (2023). Badminton self-training system based on virtual reality. In 2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) (pp. 1659-1663). https://doi.org/10.1109/APSIPAASC58517.2023.10317490
[12]Wang, F., et al. (2017). Residual attention network for image classification. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 6450-6458). Honolulu, HI, USA. https://doi.org/10.1109/CVPR.2017.683
[13]Canny, J. (1986). A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 8(6), 679-698. https://doi.org/10.1109/TPAMI.1986.4767851
[14]Hart, P. E. (2009). How the Hough transform was invented [DSP History]. IEEE Signal Processing Magazine, 26(6), 18-22. https://doi.org/10.1109/MSP.2009.934181
[15]Shimrat, M. (1962). Algorithm 112: Position of point relative to polygon. Communications of the ACM, 5(8), 434. https://doi.org/10.1145/368637.368653
[16]Hormann, K., & Agathos, A. (2001). The point in polygon problem for arbitrary polygons. Computational Geometry, 20(3), 131-144. https://doi.org/10.1016/S0925-7721(01)00012-8
[17]Hartley, R., & Zisserman, A. (2003). Multiple view geometry in computer vision. Cambridge University Press.
[18]Goodfellow, I. (2014). Generative adversarial networks. arXiv preprint arXiv:1406.2661.
[19]Nvidia. (2016). What’s the difference between artificial intelligence, machine learning and deep learning?. Retrieved from https://blogs.nvidia.com/blog/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
[20]LeCun, Y., Boser, B. E., Denker, J. S., Henderson, D., Howard, R. E., Hubbard, W. E., & Jackel, L. D. (1989). Backpropagation applied to handwritten zip code recognition. Neural Computation, 1, 541-551.
[21]Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735-1780. https://api.semanticscholar.org/CorpusID:1915014
[22]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 779-788). https://doi.org/10.1109/CVPR.2016.91
[23]WongKinYiu. (2024). YOLOv9: Implementation of paper - YOLOv9: Learning what you want to learn using programmable gradient information. Retrieved from https://github.com/WongKinYiu/yolov9
[24]Toshev, A., & Szegedy, C. (2014). DeepPose: Human pose estimation via deep neural networks. In 2014 IEEE Conference on Computer Vision and Pattern Recognition (pp. 1653-1660). https://doi.org/10.1109/CVPR.2014.214
[25]Google. (2019). Mediapipe. Retrieved from https://github.com/google-ai-edge/mediapipe
[26]Cao, Z., Hidalgo, G., Simon, T., Wei, S. -E., & Sheikh, Y. (2021). OpenPose: Realtime multi-person 2D pose estimation using part affinity fields. IEEE Transactions on Pattern Analysis and Machine Intelligence, 43(1), 172-186. https://doi.org/10.1109/TPAMI.2019.2929257
[27]Gates, B. (1995). The road ahead. New York: Viking.
[28]Ashton, K. (1999). That internet of things. RFID Journal, 22(1), 1-7.
[29]IoT Analytics. (2024). State of IoT 2023: Number of connected IoT devices growing 16% to 16.7 billion globally. Retrieved from https://iot-analytics.com/number-connected-iot-devices/
[30]ETSI. (2013). Machine-to-Machine communications (M2M); Functional architecture. ETSI TS 102 690 V2.1.1 (2013-10).
[31]陳運家, 洪友廉, 陳廣熙, 陳彩霞. (2000). 羽毛球教學手冊. Retrieved from https://resources.hkedcity.net/resource_detail.php?rid=643598345