Bennetts, Mark (2024) Quality evaluation of the Theia3D markerless motion capture system for measuring football-specific movements. Research Master thesis, Victoria University.

Abstract

The evolution of limb-tracking systems in elite football (soccer) has facilitated enhanced player technique analysis, load quantification, injury risk identification, and referee decision-making assistance. Three-dimensional (3D) markerless motion capture systems provide a non-invasive alternative to traditional laboratory-confined 3D marker systems, serving as a feasible tool for assessment of football specific limb kinematics in-game. However, limb-tracking in football has typically focused on officiating support in-game, with the quality of spatiotemporal limb- tracking data captured by markerless motion capture systems for use in training and tactical decision making currently not reported. As such, this study aims to assess the magnitude of difference between the ‘gold standard’ Vicon motion capture system and a Theia3D markerless motion capture system during football specific movements (i.e., dribbling, kicking, tackling, limb tracking during player-dense phases of play). This builds on previous research by introducing limb kinematics assessment into a football setting, expanding from a traditionally clinical context. This thesis enhances the understanding of markerless motion capture technology in a football specific context to provide guidelines on the integration and use of markerless motion capture systems in elite football environments. Three males and one female, with a mean age of 23 ± 0.7 years, an average height of 179.9 ± 5.7 cm, and a mean weight of 80.1 ± 3.3 kg participated in various football specific movements while being recorded by Theia3D (v2023.1.0.3161) and Vicon (2.15 and 2.13) systems. Acquired movement data was exported as .csv files and run in RStudio v2024.04.2-764 whereby it was cleaned, frame synchronised, aligned, and produced as interpretable data. The Theia3D system produced overall MAE and RMSE values of 0.19 ± 0.06 m and X-axis: 84.8 ± 83.6°, Y-axis: 18.4 ± 15.6°, Z-axis: 42.9 ± 31.6°, respectively. The Theia3D system proved to be effective at analysing isolated movement, with low magnitude of difference between the two systems during shooting, passing, and dribbling. The system produced high magnitude of difference during congested, multi-player trials, with misidentification of players inflating error values, suggesting end-users would most benefit from this technology by applying it to isolated skill assessment for technique analysis, injury identification and prevention, and talent identification, rather than team tactical analysis and officiating assistance in match play. Future research may be directed towards full body analysis, outdoor capture volume, more detailed post-capture filtering, and assessing alternative kinematic metrics (i.e., angular velocity).

Search Google Scholar

Repository staff login