Abstract
Abstract
Soccer shooting relies on instantaneous explosive movements and coordinated multi-joint force production, whereby aberrant movement patterns may contribute to the development of cumulative lower extremity injuries. Traditional biomechanical analyses and image-based motion evaluation approaches are frequently limited by costly equipment, poor operational flexibility, and subjective manual assessment procedures. In this study, a Python-based analytical framework was developed based on the YOLO pose estimation algorithm. Through model fine-tuning on soccer-specific shooting motion datasets, the established framework achieves automatic detection of lower limb keypoints and quantitative calculation of joint angles during shooting movements. Three core phases of the shooting motion, namely backswing preparation, ball impact, and follow-through termination, were selected for comparative analysis of hip, knee, and ankle angle kinematics between standard and atypical shooting postures. We further investigated the compensatory force production characteristics induced by aberrant movements to evaluate lower extremity injury risk, based on which a three-tier injury risk warning system was constructed. The results showed that the main features of atypical shooting postures are too limited hip extension, too much knee extension and an early compensatory activation of the knee joint. These kinematic deviations affect the force transmission through the hip–knee–ankle kinetic chain, lead to increased mechanical loading of the local joints, and may further contribute to the risk of hamstring strain, patellofemoral pain and ankle sprain. Specifically, the proposed method only needs a common monocular camera to preliminarily screen the risk of motion, providing a feasible technical solution for soccer grassroots training in motion evaluation and preventive management of lower extremity injuries. There are some caveats in this research study. First, the findings might not hold true in general because of monocular imaging projection errors, a relatively small sample size, and limited dimensionality of the features. In the future, the three dimensional kinematic parameters and multi-modal physiological data will be added to the injury risk assessment model to further improve the robustness and prediction accuracy.