Landmine Press Kinematics Measured with an Enhanced YOLOv8 Model and Mathematical Modeling

Area of Science:

• Biomechanics
• Computer Vision
• Sports Science

Background:

• The landmine press is a key upper-body strength assessment, but current monitoring tools like linear position transducers (LPTs) have limitations.
• LPTs require physical attachment, use proprietary software, and can be inaccurate under high loads due to sensor drift and noise.

Purpose of the Study:

• To develop and validate a markerless, non-contact, vision-based system for measuring landmine press kinematics.
• To overcome the constraints of traditional LPTs in assessing upper-body push strength.

Main Methods:

• Developed a vision-based system utilizing an enhanced YOLOv8-OBB model with a polarized self-attention mechanism, C3k2 module, and SPPF structure.
• Integrated a mathematical modeling framework to calculate kinematic indicators during the landmine press concentric phase.
• Compared the vision system's measurements against a linear position transducer (LPT) device (GymAware) across various loads (20-35 kg) in 247 trials.

Main Results:

• The vision system achieved high detection accuracy (mAP@0.5 of 0.995) for barbell targets.
• Strong correlations (r > 0.85) were found between the vision system and LPT for peak velocity, mean velocity, peak power, and mean power.
• The vision system showed comparable stability to LPTs under high loads and demonstrated predictable velocity overestimations.

Conclusions:

• The developed vision-based system provides a reliable and practical alternative for monitoring landmine press kinematics.
• This non-contact method offers advantages over LPTs, particularly in mitigating sensor-related errors under high-load conditions.
• The system is suitable for both strength training applications and scientific research requiring accurate kinematic data.