北京工业职业技术学院学报

深度学习方法在计算机视觉方面成绩优异。为克服传统羽毛球动作评估方式主观性强、效率低以及缺乏精准量化分析等不足，满足运动员对精准、高效训练反馈的需求，设计了羽毛球动作标准化评估系统。系统将深度学习与人体姿态估计技术相结合，采用基于OpenPose框架的人体姿态估计技术，利用人体关键点信息，将羽毛球运动员动作细分为扣球、防御、反手和落网回击，并与标准动作进行比对，对运动员训练动作进行评估打分。实验证明：系统在羽毛球动作评估方面效果良好，运动员可以充分利用该系统规范动作，提高比赛成绩。

Deep learning methods have achieved remarkable success in the field of computer vision. To address the shortcomings of traditional badminton movement assessment methods, which are highly subjective, inefficient, and lack precise quantitative analysis, meeting athletes' demand for accurate and efficient training feedback, a standardized evaluation system for badminton movements was designed. The system integrates deep learning with human pose estimation techniques, employing the OpenPose framework to extract human key points. It categorizes badminton players' movements into four types: smash, defensive, backhand, and net shot recovery, then evaluates training actions by comparing them with standard movements through scoring. Experimental results demonstrate the system's effectiveness in badminton movement assessment, enabling athletes to optimize their techniques through standardized training and enhance competitive performance.