查詢結果分析
來源資料
頁籤選單縮合
| 題 名 | 影像動作辨識系統用於羽球競賽對情蒐工作的效益:系統性回顧=Effect of Video Action Recognition System on Badminton Competition Information: A Systematic Review |
|---|---|
| 作 者 | 陳羿揚; | 書刊名 | 中華體育季刊 |
| 卷 期 | 38:2 2024.06[民113.06] |
| 頁 次 | 頁131-150 |
| 分類號 | 312.831 |
| 關鍵詞 | 深度學習; 卷神經網路; 穿戴裝置; 運動學; 自動化; Deep learning; Convolutional neural network; Wearable device; Kinematics; Automation; |
| 語 文 | 中文(Chinese) |
| DOI | 10.6223/qcpe.202406_38(2).0003 |
| 中文摘要 | 傳統羽球情蒐皆在比賽現場或藉由錄像回放採用徒手記錄簡單參數,有效提供教練團對戰組合 的對手競賽習性;但也有著需耗費大量時間、人力與可能因為分心而誤判等劣勢。如今以深度學習、 穿戴裝置與攝影機製作影像動作辨識系統,運用於羽球情蒐與戰術策略逐漸盛行。本研究目的透過 系統性回顧深度解析 2019 年 9 月至 2023 年 9 月共 4 年間,影像動作辨識系統用於辨識羽球動作姿 態需採用何種深度學習模式、基於哪種提取法、參考哪些數據資料庫、適用於哪種層級運動員、能 分析多少影片事件數量、影片分辨率、幀率範圍多寡,具體能觀測哪些事件 (動作姿態)、辨識召回 率、準確率與識別率為何;並提出實務應用、未來研究方向與相關建議,供羽球教練團、情蒐人員, 深入瞭解動作辨識系統應用於羽球競賽輔助情蒐的具體效益。 |
| 英文摘要 | Traditional methods of gathering badminton information involved manually recording simple parameters at the game site or reviewing video footage multiple times. These methods effectively provided the coaching team with insights into the opponent’s competitive habits. However, they were time-consuming, required significant manpower, and were prone to errors due to distractions. In recent years, image motion recognition systems that utilized deep learning, wearable devices, and cameras had gained popularity in badminton information gathering and tactical strategy development. These systems offered a more efficient and accurate alternative to traditional methods. This study systematically reviewed empirical articles written in both Chinese and English from September 2019 to September 2023. The discussion that followed was based on several key considerations: the type of deep learning model suitable for recognizing badminton movement postures, the appropriate extraction method, the database to refer to, the athletes to apply the model to, the number of video events that could be analyzed, the video resolution, the frame ranges, and the types of events (movement postures) that could be observed. The study also evaluated the recognition recall, accuracy, and recognition rates. It further discussed the practical applications of motion recognition systems in badminton teaching, training, and intelligence gathering, and suggested future research directions. The aim was to provide badminton coaches and intelligence collectors with a better understanding of the specific benefits of applying motion recognition systems to badminton competition intelligence collection. The study concluded with practical applications, future research directions, and related recommendations. |
本系統中英文摘要資訊取自各篇刊載內容。