早期籃球技能訓練與爭議球判定皆以教練與裁判主觀經驗為依據,如今影像動作辨識系統與穿戴裝置,陸續被提倡用於量化動作姿態、辨別專項技能優劣與輔助吹判,期望能更客觀提升整體產業價值與吹判公正性,然而動作辨識系統如何應用於辨識籃球動作技能與輔助判定爭議球,仍缺乏有系統的彙整近年相關實證研究釐清具體效益。本研究目的為透過系統性回顧釐清2019年1月至2022年8月,共3年7個月間,影像動作辨識系統用於區分籃球動作技能與輔助判定爭議球的研究現況。並且解析此類主題適用的對象(參與者)、使用數據資料庫、觀測哪些事件、影片分辨率與幀數、影片事件數量與準確度為何。接續提出此類主題未來研究方向與相關建議,供相關研究員與賽事從業人員,深入瞭解動作辨識系統應用於籃球訓練與輔助吹判的具體效益。
Early basketball training and disputed ball judgments mostly were based on the personal experience of coaches and referees. Nowadays, video action recognition systems and wearable devices have been successively promoted to quantify movement postures, distinguishing specialized skill levels, and assist in adjudication. We expecting the value of the technology industry would be increased objectively as well as the fairness of the judgment. However, there is still a lack of systematic review regarding the application and feasibility of how video action recognition systems could be used to identify basketball movement skills and assist in determination of disputed balls in recent empirical studies. This study systematically reviewed empirical and reviewed articles written in Chinese and English ranged from January, 2019 to August, 2022. The follow-up discussion was based on the deep learning models, classification strategies, participants observed, the use of a database, what type of events being observed, video resolution and frame rate, and what is the number of video events and how accurate are they? The results provided an insight of the effectiveness of video action recognition systems in basketball training and referee assisting for researchers and competition officials.