Concurrent validity of markerless motion capture for occupational lifting: Kinematics and ergonomic risk assessment

Abstract

Markerless motion capture has the potential to enhance ergonomic risk assessment, but evidence remains limited for occupational manual handling tasks of increasing complexity. This study compared a multi-view system (OpenCap) and a monocular system (MediaPipe) against a marker-based reference system (Vicon) for lifting kinematics. Outcomes were assessed using the Rapid Entire Body Assessment (REBA) risk assessment tool. Twenty healthy participants performed three two-handed lifting tasks of increasing complexity under controlled laboratory conditions. Errors increased by 15–30% from symmetrical to asymmetric multi-planar tasks, with both systems demonstrating lower error in the sagittal plane (RMSE = 9.9–31.2°) than in the frontal and transverse planes. Both systems were reliable for symmetrical lifting, including high REBA agreement, but validity declined for tasks involving trunk rotation and walking. MediaPipe showed lower error for trunk and proximal joints, whereas OpenCap performed better for distal kinematics, informing practical system selection.