Effectiveness of an automatic tracking software in underwater motion analysis.

Area of Science:

• Biomechanics and Sports Science
• Computer Vision and Image Analysis
• Aquatic Sports Performance

Background:

• Kinematic analysis using marker tracking is crucial in sports science for athletes and coaches.
• Existing automatic tracking software is not optimized for underwater motion analysis.
• Underwater environments present unique challenges like low image quality and target occlusion.

Purpose of the Study:

• To evaluate the effectiveness of a newly developed software (DVP) for automatic underwater motion tracking.
• To compare the DVP software's performance against a commercial tracking software (COM).
• To assess the degree of automation by measuring required manual interventions.

Main Methods:

• Developed DVP software utilizes the Kanade-Lucas-Tomasi feature tracker for automatic marker tracking.
• Manual tracking of 2940 marker positions across 21 aquatic exercise videos served as the ground truth.
• Automatic tracking was performed using DVP and COM software, with manual corrections triggered if marker position error exceeded 4 pixels.

Main Results:

• The DVP software required significantly fewer manual interventions (7.4%) compared to the commercial software (17.8%).
• DVP demonstrated lower manual intervention rates across different exercise modes (5.6%–29.3% reduction) and marker types (9.9% reduction).
• The results indicate a higher degree of automation for the DVP software in underwater kinematic analysis.

Conclusions:

• The developed DVP software is a valid and effective tool for automatic underwater motion analysis.
• DVP offers improved efficiency in kinematic analysis for swimming and other aquatic sports due to its high automation.
• The software's feature-tracking algorithm successfully addresses challenges in underwater environments, reducing the need for manual supervision.