Accuracy of a markerless system to estimate the position of taekwondo athletes in an official combat area

Obtaining kinematic analysis through a gold-standard method on sports during official competitions is challenging. Markerless tracking system can be an alternative to provide physical and technical-tactical performance data about the athletes. Therefore, we aimed to verify whether the markerless system (OpenPose) accurately tracks taekwondo athletes' positions on the official combat area. Six taekwondo athletes performed steps and kicks in a self-selected way in two conditions: with and without an opponent. To capture the movement, seven optoelectronic (marker-based system) and two digital cameras (OpenPose markerless system) were placed around the combat area at an acquisition frequency of 120 Hz and 30 Hz, respectively. Positions of the body centre of mass and midpoint of the feet were calculated and analysed through two and three-dimensional reconstruction both. The root mean square (RMS) error comparing both methods ranged from 0.13 to 0.32 m and 0.05 to 0.16 m for two and three-dimensional analysis, respectively. Bland-Altman analysis accepted the agreement between the capture systems, and the intraclass coefficient correlation (ICC) values were classified as excellent (ICC > 0.90) in all analysis. In conclusion, the OpenPose markerless system presents promising results for its use in tracking taekwondo athletes' position on the competition area.