Research on Image Recognition Technology to Assess the Quality of Badminton Players' Stroke Action and Training Improvement Strategy

Abstract

This paper establishes a motion agent system based on image recognition to assess the quality of badminton players' stroke techniques, and provides personalized training improvement plans using deep learning and biomechanical analysis. Infrared dot high-speed cameras are used to capture the movements of the athlete's shoulder joint, elbow joint, and wrist joint, among other positions. Convolutional neural networks and long short-term memory networks are employed to classify and evaluate the movements of forehand strokes, pushes, and hooks. The action recognition accuracy rate and feature point detection error rate are 96.2% and 93.2%, respectively. The athletes' technical consistency, movement standardization, and training effectiveness improved by an average of over 30% compared to pre-training levels. This paper effectively overcomes the subjectivity and errors of traditional evaluation methods, providing a quantitative, scientific, and accurate method for evaluating badminton technical movements. This study has practical application value and provides theoretical and methodological guidance for the use of image recognition in research on other sports.