Indoor bouldering exposes climbers to falls whose conditions remain difficult to characterize. This thesis evaluates an approach combining markerless 3D reconstruction (Pose2Sim), inertial measurements, and finite element simulation to analyze landings in athletes.
The protocol produced 225 measurements in 22 athletes. The article focuses on 40 falls performed by 10 athletes aged 14 to 17 years, recorded with five action cameras (GoPro HERO12, GoPro, USA; 2.7K, 240 fps) and three IMUs (Blue Trident, Vicon, UK; ±200 g, 1600 Hz). Pose2Sim was compared with Kinovea and IMUs. Mean segmental errors ranged from -0.5% to +8.9%; ICCs with Kinovea ranged from 0.958 to 0.989 for fall height and peak velocity, but Pose2Sim underestimated peak accelerations at the ankle and sacrum.
A reference fall was used to build an exploratory simulation with THUMS AM50 V7.1 in LSDYNA. Differences between simulated and experimental accelerations ranged from -24.9% at the right ankle to +8.8% at the right sacrum. The maximum equivalent von Mises stress in the cortical bone of the tibias reached 106.9 MPa, within the longitudinal compressive yield-stress range reported by Morgan et al. (2018). Maximum strains in the medial (22.5%) and lateral (39.6%) complexes overlapped the partial tear (21.3% to 29.7%) and complete rupture (28.0% to 35.6%) ranges reported by Button et al. (2013).
Markerless video characterizes global kinematics, IMUs remain necessary for impact analysis, and simulation explores internal mechanical loading. However, these results are limited to one non-injurious reference fall and do not constitute a clinical injury prediction. Perspectives include exploiting the complete dataset, adapting the models to the experimental conditions, and studying more varied scenarios.
Carretier, N. (Author),
Clément (Supervisor) &
Beauséjour (Co-supervisor),
29 Jun 2026Student thesis: Master's thesis › Master in Engineering: Engineering