以視覺為基礎之撞球自我訓練系統

Vision-based Self-training System for Billiards

蔡厚群(Hou-Chun Tsai)

指導教授 ： 李素瑛 ； 蔡文錦 ； 陳華總

國立交通大學/資訊學院/資訊科學與工程研究所/碩士(2016年)

摘要

正確的選取球桿擊球的角度對於撞球的初學者來說無疑是一項挑戰。現存大多數以視覺為基礎的撞球訓練系統仰賴於球桌正上方的影像擷取設備，其分析環境設置對於一般使用者來說並不實際。本論文提出了基於電腦視覺的撞球自我訓練系統，目的在於結合使用者所拍攝的多張球桌側視圖，以建立俯視的球桌模擬圖，再搭配由使用者視角所拍攝的球桿擊球角度，提供使用此擊球角度擊球之後母球以及目標球的移動軌跡預測，軌跡視覺化後回傳給使用者作為訓練指示。首先，我們對三張側視圖分別進行球桌擷取、球輪廓擷取和球桿角度估測，以便結合各張側視圖中的球並且投影定位各球在模擬圖上的位置，最後再搭配使用者另外拍攝的球桿角度圖，估算俯視圖上的球桿擊球角度。除此之外,我們也模擬母球以及目標球在撞擊之後各自的移動軌跡，並且透過視覺化提供使用者指示。我們的系統中，智慧眼鏡同時扮演了球桌影像拍攝以及預測軌跡顯示的角色，增加了系統實用性之餘，也能提供使用者更加直觀的訓練指示與協助。

關鍵字

運動影像分析 ； 運動訓練 ； 球桿角度估算 ； 多媒體應用

並列摘要

Precisely selecting the aiming angles is undoubtedly a challenge to inexperienced billiard players. However, most studies focusing on self-training systems for billiards are based on ceiling-mounted cameras, which are not practical to general users. As a result, we propose a vision-based self-training system for billiards, named “Improve My Shot,” to help inexperienced billiard players improve their shots. The system can generate top-view simulation table by integrating the side-view images from three or more angles and predict the trajectories of both the cue-ball and the object-ball after collision according to the additional images with the cue-sticks captured from intelligent glasses. The system consists of three main processing modules, including table contour extraction, ball contour extraction and cue-stick direction estimation. Each ball can be located in the real-world coordinates by integrating ball projections from different angles and can be visualized in the top-view simulation table. Finally, with the estimated cue-stick direction, we can predict the trajectories of both the cue-ball and the object-ball after collision. The trajectories are then visualized and presented on the intelligent glasses to provide users with instructions. In the proposed system, Google Glass is adopted for image capturing and result display. After users capture images via their Google Glass, the images are sent to the server for processing and the resulting images with the instruction lines are sent back for display afterwards, which not only enhances the practicability of the system but also helps billiard players improve their shot in more intuitive way.

並列關鍵字

sports video analysis ； sports training ； cue-stick direction estimation ； multimedia application

參考文獻

[1] M. H. Hung and C. H. Hsieh, “Event detection of broadcast baseball videos,” IEEE Transactions on Circuits and Systems for Video Technology, vol.18, no.12, pp. 1713–1726, 2008.

[2] Z. Wang, J. Yu, and Y. He, “Soccer video event annotation by synchronization of attack–defense clips and match reports with coarse-grained time information,” IEEE Transactions on Circuits and Systems for Video Technology, [Online] available: IEEE Xplore, doi: 10.1109/TCSVT.2016.2515280, 2016.

[3] A. Javed, K. B. Bajwa, H. Malik, and A. Irtaza, “An efficient framework for automatic highlights generation from sports videos,” IEEE Signal Processing Letters, vol.23, no.7, pp. 954-958, 2016.

[5] M. C. Hu, M. H. Chang, J. L. Wu, and L. Chi, “Robust camera calibration and player tracking in broadcast basketball video,” IEEE Transactions on Multimedia, vol. 13, no. 2, pp. 266-279, 2011.

[9] H. C. Miles, N. Musembi, S. R. Pop, and N. W. John, “Virtual environment for rugby skills training,” Proceedings of Joint VR Conference of EuroVR and EGVE, pp. 64-66, 2011.