Marker based and markerless motion capture for equestrian rider kinematic analysis: A comparative study.

Overview

• This study compared marker-based and markerless motion capture systems to evaluate their accuracy in capturing equestrian rider movements.
• It aimed to determine if markerless technology could provide similar joint and segment angle data as traditional marker-based methods during walking and trotting on a horse.

Introduction and Objective

• Traditional marker-based motion capture is the standard for analyzing biomechanical movements but requires physical markers attached to the subject.
• Markerless systems, which use video data without markers, offer a less intrusive and more flexible alternative but their accuracy in equestrian contexts is unclear.
• The study hypothesized that markerless motion capture could yield comparable kinematic data to marker-based systems for riders.
• The objective was to compare joint and segment angles recorded during walking and trotting with both systems, focusing on key joints and body segments relevant to riding biomechanics.

Methodology

• Participants: 10 healthy adults acting as equestrian riders.
• Trial Setup:
  • Each participant completed five dynamic trials during both walking and trotting gaits while mounted on a horse.
  • A 12-camera marker-based system and an 8-camera 2D video-based markerless system were simultaneously used and synchronized.
• Data Collected:
  • 3D joint angles for the hip, knee, shoulder, and elbow.
  • Global trunk and pelvis angle measurements.
• Data Analysis:
  • Root mean square difference (RMSD) was calculated across gait cycles to quantify error between systems.
  • Statistical parametric mapping (SPM) paired t-tests were applied to assess significant differences across the gait cycle phases.

Results

• RMSD findings:
  • The sagittal trunk angle showed the lowest average RMSD of 2.0°, indicating close agreement between systems.
  • The elbow rotation had the highest RMSD at 19°, indicating greater discrepancy.
  • Error magnitudes were similar for both walking and trotting conditions.
• SPM Analysis:
  • During walking:
    • Significant increases in hip flexion were detected by the markerless system throughout the gait cycle (p < 0.001).
    • Elbow flexion was significantly higher in specific gait cycle intervals (24-47% and 63-100%, with p-values of 0.03 and < 0.001).
  • No significant differences were found in measurements within the transverse plane, although RMSD values tended to be higher suggesting less consistent accuracy there.

Discussion

• Potential causes for discrepancies include:
  • Limited joint range of motion in medial limbs when mounted, affecting markerless tracking accuracy.
  • Obscuration of body parts by the horse or rider apparel causing markerless detection challenges.
• The higher errors and less consistency in the transverse plane suggest caution when interpreting such data from markerless systems in equestrian contexts.
• Despite these challenges, markerless systems demonstrated good performance for several joint angles, especially in the sagittal plane.

Conclusions and Implications

• Markerless motion capture shows promise as a practical alternative for rider movement analysis with potential applications in rider biofeedback.
• The suitability of markerless data depends on:
  • The specific joint or segment being analyzed.
  • The acceptable margin of error for the intended use case.
• This technology may reduce some practical limitations of marker-based systems, such as marker placement and interference from motion, but is less reliable for certain planes of movement and specific joint rotations.
• Future work should focus on refining markerless algorithms, improving data capture in occluded or complex riding environments, and validating further movement conditions.