Assessing Agreement and Usability of Kinematic Measures between Markerless and Marker-Based Motion Capture Systems

Abstract

Markerless motion capture technology can improve efficiency and accessibility in human biomechanics research, yet its merits as an accurate and precise measurement tool are not fully vetted. As markerless technology relies on 2D video-based images from cameras, the effects of different camera positions on reliability of resulting kinematic data is not well understood. We co-recorded various movement tasks from 11 participants using three independent camera systems: two identical markerless camera systems positioned at different heights to assess precision, and a marker-based camera system to assess accuracy, focusing on lower limb kinematics. Both novice and experienced motion capture users assessed usability of the markerless system. When assessing precision, the markerless system produced a mean intraclass correlation coefficient of 0.95±0.01 and mean root mean square error of 1.91±0.98° across different joint angles and tasks. Relative to a marker-based system, the markerless system had a mean Lin’s concordance correlation coefficient of 0.51±0.19 and mean root mean square error of 3.29±2.36°, after making bias-adjusting shifts in joint angles. Time-series curve comparisons also showed minimal differences between both marker-based and markerless systems. Users rated the markerless system highly for ease and confidence of use. While differing definitions for the pelvis and hips may systematically affect comparisons between marker-based and markerless systems, our data indicates that the evaluated markerless motion capture technology demonstrated good precision between identical systems using different camera perspectives, was accurate relative to a standard marker-based system, and was generally easy for novices and experienced users to acquire and process data.

Keywords: Markerless Motion Capture, Sports, Validation, Usability, Biomechanics, Kinematics