全球觀看籃球比賽的人數總計約超過22億人，根據外國媒體Sports Show在2020年公布全球最受歡迎的運動賽事，籃球在所有球類中排名第三，可看出籃球是一項非常熱門的運動。近年來運動分析的研究相當熱門，透過將生成對抗網路應用在籃球領域能夠幫助球隊提升籃球攻防戰術的素養，開發出基於深度學習之籃球攻防戰術軌跡生成系統。本系統開發目的為進攻球隊使用者在分析研究防守球隊可能會出現的防守方法時，通常只能使用經驗判斷推測，若透過本系統自動產生防守戰術軌跡供進攻球隊參考，進攻球隊可更加理解實戰中可能會遇到的防守戰術，可提升球員的戰術素養讓球隊提早思考應對方法。
本系統透過使用者將一段真實籃球比賽攻防片段輸入，系統主要分為兩個子系統：投影轉換子系統與防守戰術軌跡生成子系統。投影轉換子系統主要分為三個步驟，第一為球場上球員與球的偵測方法，接著界定球場的範圍。第二為場上球員分隊使用球衣顏色做為辨別的依據。接著為3D球員座標投影計算出單應矩陣將對應的3D座標映射在2D戰術板球場座標系中並記錄為檔案作為防守戰術軌跡生成子系統的輸入。最後一個步驟使用生成對抗網路來進行防守戰術軌跡生成。
本研究實驗結果顯示，透過影像處理得到球場邊線同時界定新的球場範圍可有效省略透過觀察手動決定球場頂點的步驟，減少時間成本。加入球員分隊的功能計算該區域內的色調特徵與顏色強度特徵，使用K-means clustering 將該二類特徵將場上球員分成兩隊，以利最後映射至平面戰術板座標系還原出真實比賽的情況。映射結果的球員正確率達到了77.2%，籃球則為61.0%。本系統結合了真實籃球比賽片段與防守戰術軌跡生成系統產生虛擬的防守戰術軌跡。

Basketball has a vast global audience of 2.2 billion, ranking as the third most popular sport. Recent sports analysis research leverages Generative Adversarial Networks (GANs) to develop a deep learning system for generating basketball offensive and defensive tactics. This system's aim is to provide offensive teams with auto-generated defensive tactics trajectories, improving understanding of potential defensive strategies, enhancing players' skills, and enabling more effective team strategies.

The system has two main subsystems: the Projection Transformation Subsystem and the Defensive Tactics Trajectories Generation Subsystem. The Projection Transformation Subsystem involves three steps: player and ball detection on the court, defining court boundaries, and distinguishing players based on jersey colors. 3D player coordinates are projected onto a 2D tactical board coordinate system via homography, and this data is used for the Defensive Tactics Trajectories Generation Subsystem, which employs Generative Adversarial Networks to generate defensive tactics trajectories.

Experimental results indicate that image processing for court boundaries and player teams significantly reduces time costs. The inclusion of the player team function enables calculation of region color and intensity features, with K-means clustering dividing players into teams for mapping to a flat tactical board coordinate system, replicating real game scenarios. Player mapping accuracy reached 77.2%, while basketball detection accuracy was 61.0%. This system combines real basketball game segments with the defensive tactics generation system to create virtual defensive tactics trajectories.

Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998.
R. Girshick, J. Donahue, T. Darrell and J. Malik, “Region-Based Convolutional Networks for Accurate Object Detection and Segmentation,” Proceedings of the IEEE, vol. 38, no. 1, pp. 142-158, 1 Jan. 2016.
R. Girshick, “Fast R-CNN,” Proceedings of the IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, pp. 1440-1448, 2015.
S. Ren, K. He, R. Girshick and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, 1 June 2017
K. He, G. Gkioxari, P. Dollár and R. Girshick, “Mask R-CNN, ” Proceedings of the IEEE International Conference on Computer Vision (ICCV), Venice, Italy, pp. 2980-2988, 2017.
T. Y. Lin, P. Dollar, R. Girshick, K. He, B. Hariharan, and S. Belongie, “ Feature Pyramid Networks for Object Detection,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2117-2125, 2017.
K. He, X. Zhang, S. Ren and J. Sun, “Deep Residual Learning for Image Recognition,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp. 770-778, 2016.
S. Janssen (2019, Dec. 21). “The journey towards creating a Basketball mini-map” Availble: https://www.linkedin.com/pulse/journey-towards-creating-basketball-mini-map-stephan-janssen (Dec. 21, 2019).
G. Mohan (2021, Dec. 24). “Developing a Basketball Minimap for Player Tracking using Broadcast Data and Applied Homography” Availble: https://medium.com/mlearning-ai/developing-a-basketball-minimap-for-player-tracking-using-broadcast-data-and-applied-homography-433183b9b995 (Dec. 24, 2021).
H.T. Nguyen, H.S. Hoang, T.P.D. Chu, Q.D. Vu, and M.H. Luu,“A video-based tracking system for football player analysis using Efficient Convolution Operators,” Proceedings of 2019 IEEE Conference on Advanced Technologies for Communications (ATC), Hanoi, Vietnam, 2019, pp. 149-154.
C. Y. Chen, W. Lai, H. Y. Hsieh, W. H. Zheng, Y. S. Wang, and J. H. Chuang, “Generating Defensive Plays in Basketball Games,” Proceedings of the 28th ACM International Conference on Multimedia (MM ’18), 9 pages, New York, 2018.
H. Y. Hsieh, C. Y. Chen, Y. S. Wang, and J. H. Chuang, “Basketball GAN: Generating Basketball Play Simulation Through Sketching,” Proceedings of the 27th ACM International Conference on Multimedia (MM ’19), 9 pages, New York, 2019.
A. Miller, “Possession Sketches: Mapping NBA Strategies,” Proceeding of 2017 Sport Analytics Conference, Hynes Convention Center, 2017.
A. Nistala, and J. Guttag, “Using Deep Learning to Understand Patterns of Player Movement in the NBA,” Proceedings of 2019 MIT SLOGN Sport Analytic Conference, Boston, 2019.
L. Javadpour, J. Blakeslee, M. Khazaeli, and P. Schroeder, “Optimizing the best play in basketball using deep learning,” in Journal of Sports Analytics, vol. 8, no. 1, pp. 1-7, 2022.
C. H. Chen, T. L. Liu, Y. S. Wang, H. K. Chu, N. C. Tang, and H. Liao, “Spatio-Temporal Learning of Basketball Offensive Strategies,” Proceedings of the 23rd ACM international conference, Multimedia, 13 October 2015.
Y. Liu, P. Sun, N. Wergeles, Y. Shang, “A survey and performance evaluation of deep learning methods for small object detection” Proceedings of Electrical Engineering and Computer Science (EECS), USA,2021.
Y. Yoon, H. Hwang, Y. Choi, M. Joo, H. Oh, I. Park, K. Lee, J. Hwang, “Analyzing Basketball Movements and Pass Relationships Using Realtime Object Tracking Techniques Based on Deep Learning” Proceedings of IEEE Access, vol. 7, pp. 56564-56576, 2019.
L. Chen, W. Wang, “Analysis of technical features in basketball video based on deep learning algorithm,” Image Communication, Volume 83, Issue C, Apr 2020.
全球最受歡迎運動，第一意料之內，第二竟然是…？, Availble: https://mag.sportsoho.com/%E5%85%A8%E7%90%83%E6%9C%80%E5%8F%97%E6%AD%A1%E8%BF%8E%E9%81%8B%E5%8B%95
Michael Jordan, Availble: https://cdn.nba.com/manage/2021/08/michael-jordan-looks.jpg
Kobe Bryant, Availble:https://imagez.tmz.com/image/1e/4by3/2020/01/26/1e77e10e628e4ccf81a8a21b40f36d3b_md.jpg
Dwight Howard, Availble:https://slamdunkhk.com/wp-content/uploads/2022/11/Dwight-Howard-Taiw
an-Injury-2-weeks.webp
林書豪, Availble:https://cdn.ttv.com.tw/manasystem/FileData/News/23015e20-d8a2-43d6-9
714-40104c1676f0.jpg
真實籃球比賽, Availble:https://github.com/GauravMohan1/Basketball_Homography/blob/main/college_middle.png
戰術板, Availble:https://cdntwrunning.biji.co/800_b5272bba13dd94e97dff9202300e133c.jpg, https://m.media-amazon.com/images/I/61mBOXiq1+L._AC_SX679_.jpg