Practical Considerations for Effective Markerless Motion Capture Research

Abstract

In any successful research endeavour, the data collection protocol is a crucial element to consider in experiments. Tools appropriate for data collection are borne out of need, and new tools emerge as previous limitations become identified. Consequently, novel instruments and techniques undergo necessary processes of validation and testing before reaching widespread adoption. In the field of biomechanics, the most widely accepted method for analysing whole-body movements is marker-based motion capture. Limitations of marker-based motion capture, however, include marker placement errors, soft tissue artefacts, inter-operator variability, and lack of accessibility. To address these limitations, markerless motion capture technology utilizing standard video cameras and computer vision techniques offer a viable alternative for human pose estimation. This thesis examines practical considerations of markerless motion capture technology and its implications for biomechanical research. Two studies were undertaken: (i) a comparison of segment lengths, joint kinematics, and gait spatiotemporal parameter outcomes between two clothing conditions, and (ii) a pilot investigation on the impact of camera configuration, clothing color, and environment on the ability to track complex movements performed during yoga. The results of these examinations indicate that markerless motion capture is robust to changes in clothing condition and camera configuration and feasible for capturing complex movements in varied environments. Compared to marker-based motion capture systems, markerless motion capture enables high-throughput data collection in an ecologically valid manner with lower costs. It is time to challenge marker-based motion capture systems as the accepted standard of technology for collecting biomechanical information; markerless motion capture lies at the cusp of democratizing motion capture acquisition for all.