University of Twente Student Theses


The Development of a Novel Method for Measuring Biomechanical Features of Agility Using Markerless Motion Capture

Peetsma, Jasper (2023) The Development of a Novel Method for Measuring Biomechanical Features of Agility Using Markerless Motion Capture.

[img] PDF
Abstract:Assessing the agility of sports practitioners up until now has been done through agility tests and change-of-direction tests, in which the test completion time is the sole metric of performance. Within this thesis, a novel method is proposed to better understand and estimate an athlete’s agility. As time is often of the essence – especially for professional sports players and teams – the method takes practicality into account by capturing movement without needing markers or sensors to be attached to the athlete’s body. This markerless motion capture (MMC) method saves time, as only a set of two cameras needs to be set up in advance, and no extra valuable time is wasted in between the data collection of each athlete. Literature research showed correlations between agility and interlimb asymmetry, ground contact time, and take-off distance. To measure these features, OpenPose was used to detect pose keypoints in the stereo video recordings of arrowhead agility tests, which were triangulated into 3D coordinates using MATLAB. These coordinates were further developed into parameters, or key performance indicators, by calculating (differences in) joint angles, ground contact start and end times, and whole-body centre-of-mass positions, velocity and acceleration. To demonstrate how the data can be used, parameters are modelled through a generalized linear mixed-effect modelling (GLMM) approach. However, no definitive conclusions can be drawn from the model without validating the data with a gold-standard motion capture system. The work shows promising results for the use of OpenPose in the assessment of agility. The system offers many possibilities for easy collection of biomechanical data using off-the-shelf tools, even outside of lab environments. However, there is room for improvement in several steps of the process (e.g. in the camera calibration, the OpenPose configuration, the triangulation into 3D coordinates, and the joint angle calculation). These must be addressed before accurate and reliable information can be extracted.
Item Type:Essay (Master)
FC Twente, Enschede, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 54 computer science
Programme:Interaction Technology MSc (60030)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page