(10) COMPARISON BETWEEN SMARTPHONE MARKERLESS MOTION CAPTURE AND MARKER-BASED MOTION CAPTURE SYSTEMS DURING BASEBALL PITCHING

Consistent fastball velocity is an important factor for baseball pitchers. Effective kinematics, kinetics, and relative timing of segmental interactions can optimize momentum transfer and competitive outcomes.

Purpose: To compare the relationship between pitching kinematics (joint angles) from a 2D markerless motion capture application designed for a single camera smartphone to the gold standard 3D marker-based motion capture system during baseball pitching.

Methods: Fourteen Division I collegiate baseball pitchers (age = 20.9 ± 1.7 yr, height = 184.5 ± 6.6 cm, body mass = 89.1 ± 10.2 kg) volunteered for this study. Forty-eight reflective markers were attached to the body for the marker-based 12 camera 3D motion capture system (model Miqus M3; Qualisys, Goteborg, Sweden). For the markerless 2D motion capture system (ProPlayAI, Ontario, Canada), the video was uploaded to PitchAI’s software where the program performed a 2D pose estimation of 19 joint centers from a single camera smartphone. PitchAI then transformed the 2D data into a 53-marker 3D joint center model. The first fastball recorded by both PitchAI and Qualisys was used for analyses. Each pitch was adjusted to begin from the peak vertical knee position and end at ball release. Discrete-time points (foot plant, maximum external rotation, and ball release) were evaluated for 14 different joint kinematics using Pearson R (r), R-Squared (r²), and paired sample t-tests to compare PitchAI’s and Qualisys data results. Statistical significance was set at an alpha level of p ≤ 0.05. Correlations were listed as high (± 0.800 - 1.0), moderately high (± 0.600 - 0.799), or moderate (± 0.400 - 0.599).

Results: Rear leg knee extension at ball release had a highly significant correlation between the two software’s (r = 0.911, r² = 0.83). A moderately high significant correlation was found for throwing arm elbow flexion at the foot plant (r = 0.734, r² = 0.54), rear leg knee extension at maximum external rotation (r = 0.608, r² =0.37), and trunk lateral tilt at foot plant (r = 0.663, r² =0.44). There was a moderately significant correlation for lead leg knee extension at ball release (r = 0.565, r² = 0.32), and a negative moderate correlation at lead leg knee extension at foot plant (r = -0.556, r ²= 0.31) and throwing arm elbow flexion at ball release (r = -0.565, r² = 0.32). Paired sample t-test identified significant differences between the two motion capture systems.

Conclusions: Similar linear-curve trends were identified throughout the full-time series of the pitching delivery for both motion capture systems. Our results indicated there were some high to moderate correlations between systems. Although there were similar trends and some correlations, paired sample t-test identified significant differences between the two technologies. PRACTICAL APPLICATIONS: PitchAI could be a useful technology to identify position-specific kinematic changes in athletes’ pitching motion, specifically for the lower extremities, lateral tilt of the trunk, and elbow flexion at foot contact and ball release. Although it could be a quicker way to access kinematic data during the pitching motion, coaches should be aware of the normative data provided for each motion capture software before making any changes to athletes’ kinematics as significant differences exist between multi-camera marker-based optical segment tracking versus markerless human modeling arising from single camera smartphone data captures.

Acknowledgements:

We would like to thank the baseball pitchers and coaching staff at Louisiana Tech University for being involved in this project.