應用深度學習YOLOv3在羽球賽事中提升偵測和辨識選手精確度之研究

A Study on Improve Precision for Detecting and Identifying Players from Badminton Games Via Deep Learning YOLOv3

蔡俊明(Chun-Ming Tsai) ； 李秋蘭(Chiu-Lan Lee)

《國教新知》 66卷1期 (2019/06) Pp. 51-64

https://doi.org/10.6701/TEEJ.201906_66(1).0004

摘要

在羽球體育界中，傳統的羽球戰術分析方式，是由教練或教練群等人力盯著羽球比賽影片來做分析，這樣做非常費時和費力，加上一直盯著羽球比賽影片看，教練和選手們的眼睛容易疲勞。因此，為了解決這些問題，我們希望提出一智慧型羽球比賽影片分析系統，來幫助教練和選手，大量且快速分析球員的戰術。在此論文中，我們主要是研究如何提升從羽球比賽影片中偵測和辨識羽球選手的精確度（Precision）。我們的方法包括採用標註整體和局部羽球選手方式，深度學習YOLOv3的訓練和測試，以及提出後處理來解決一位羽球選手同時被一個整體Bounding box和一個局部Bounding box偵測到的問題。經實驗結果，在開放測試50張影像下，精確度由50.96%提升為88.64%。

關鍵字

羽球賽事 ； 深度學習 ； 提升精確度

並列摘要

In the badminton sports world, the traditional badminton tactical analysis method is to analyze the badminton game film by the coach or coach group, which is very time-consuming and laborious. In addition, the coach and the player are always watching the badminton game. Their eyes are prone to fatigue. Therefore, in order to solve these problems, we hope to propose an intelligence badminton game film analysis system to help coaches and players to analyze the player's tactics in a large amount and quickly. In this paper, we mainly study how to improve the precision of detecting and identifying badminton players from badminton games. Our approach includes the use of an overall and partial badminton player labeling approach, training and testing by deep learning YOLOv3 and post-processing to address a badminton player being detected by both an overall bounding box and a partial bounding box. From the experimental results, the precision was increased from 50.96% to 88.64% in 50 open testing images.

並列關鍵字

badminton games ； deep learning ； improve precision

參考文獻

王志騰（2018）。利用 UWB 定位資訊進行足球運動賽事即時事件偵測。（碩士），玄奘大學，新竹市。Retrieved from https://hdl.handle.net/11296/an5w76

Redmon, J., & Farhadi, A. (2018). Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767

Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.

Schlipsing, M., Salmen, J., Tschentscher, M., & Igel, C. (2017). Adaptive pattern recognition in real-time video-based soccer analysis. Journal of Real-Time Image Processing, 13(2), 345-361.