Validation of Video-based Motion Capture for Lower Body and Trunk Kinematics with Marker Augmentation

Markerless motion capture offers a promising alternative to traditional marker-based systems by addressing limitations such as soft tissue artifacts and labor-intensive setups. This study aimed to evaluate the accuracy and reliability of the Pose2Sim framework, enhanced with Real-Time Multi-person Whole-body pose estimation (RTMW) and marker augmentation, for assessing lower body and trunk kinematics during soccer kicking. A total of 194 trials were recorded, with motion captured simultaneously by a marker-based system and markerless Pose2Sim using four action cameras. Comparisons were conducted across medial-lateral (M-L), anterior-posterior (A-P), and inferior-superior (I-S) axes. The coefficient of multiple correlation (CMC) and root mean square error (RMSE) were used to assess waveform similarity and accuracy. Strong correlations (CMC > 0.9) were observed across most joints. RMSE values showed segment and axis specific patterns, with errors below 5° for the shank, thigh, and trunk in the M-L and A-P axes, whereas larger errors were observed for the foot in the M-L and A-P axes and for the shank, thigh, and trunk in the I-S axis. These findings suggest the promise of Pose2Sim at an initial validation stage, supporting its potential as a cost-effective, open-source tool for biomechanical applications.