Purpose: This study uses OpenCV to recognise and track runners in a 1500 m race video, output their actual coordinates and calculate their actual running distance and speed. By tracking middle-distance runners, their running performance can be analysed. The traditional method of video recording and analysis for a 1500 m race can be replaced by artificial intelligence techniques. Method: This study uses a 4K high-definition (3840 × 2160 pixel resolution) camera for fixed shooting at 30 fps with a wide-angle lens and MP4 video recording format. AutoCAD was used to draw and measure the actual coordinates of the playground, and OpenCV enabled the image recognition and tracking of runners in the 1500 m competition video, output their image coordinates, calibrate the image coordinates with the actual coordinates and output their coordinate trajectory over the 1500 m distance, thereby calculating their speed in 100 m intervals. Results: Using the actual coordinates, one of the runners actual running distances was calculated as follows: the 0-300 m section was 298.9 m, the 300-700 m section was 402 m, the 700-1100 m section was 404.6 m and the 1100-1500 m section was 404.7 m. In the 1500-m race, the runner ran approximately 10 m more than the official distance. In terms of speed, we compared the effectiveness of object tracking and manual video analysis in calculating speeds per 100 m. There was no statistically significant difference between the two data sets (P = 0.23), and the difference was less than 0.5 s. Conclusion: Based on OpenCV, the identification and tracking of runners in the competition video can basically be realised, and the actual distance runners cover during the competition can be calculated. However, there are still difficulties in the object-tracking process. This study is the first to use object tracking in middle-distance running over 1500 m. Middle-distance races are more complex than sprint races, which undoubtedly present substantial object-tracking challenges. However, this new attempt will undoubtedly provide new ideas and references for the analysis of middle-distance races in the future.
© Copyright 2023 Journal of Physical Education and Sport. University of Pitesti. Alle Rechte vorbehalten.