Assisted Training Methods for Taekwondo Based on Improved OpenPose Model and Target Detection

Current training methods for Taekwondo mainly rely on the coach’s experience, the athlete’s self-perception, and the lack of objective quantitative assessment tools. To improve training efficiency and athlete performance, the study proposes an assisted training method for Taekwondo based on an improved OpenPose model and target detection. The study first optimised the OpenPose model by replacing the feature extraction network and improving the convolutional structure. Before the pose estimation, the target detection method was applied to detect the target during the training process, and then the improved OpenPose model was utilised for the pose estimation. Finally, the athlete’s movement quality was scored by combining the static and dynamic movement characteristics of training and long short-time memory. The coach also optimised and adjusted the training method according to the evaluation results. The experimental results indicated that the student satisfaction reached 92.17% after one year of application of the assisted training method proposed by the study, and the training efficiency was significantly improved. In addition, students’ test scores were improved by more than 25%. The research design method was able to make timely adjustments to the training program by evaluating the athletes’ Taekwondo training movements, thus improving the quality of training.