The aim of the present study was to assess the accuracy of a multi-camera tracking system (Mediacoach®) to track elite football players’ movements in real time. A total of 207 observations of 38 official matches from Liga 1, 2, 3™ (2nd Spanish Division, season 2017/18) were included in the study (88 defenders, 84 midfielders, and 35 attackers of the same team). Total distance (TD, m) distance in zone 4 (DZ4) at a speed of 14–21 km/h, distance in zone 5 (DZ5) at a speed of 21–24 km/h (DZ5), distance in zone 6 (DZ6) at a speed of ≥24 km/h, maximum speed (km/h), and number of sprints (actions above 24 km/h) were registered with the Apex® GPS system (STATSports™, Newry, N. Ireland) and Mediacoach® semi-automatic tracking system (LaLiga™, Madrid, Spain). The level of agreement between variables estimated by the two systems was analyzed. Bias was also calculated by deducting the GPS estimated value from the video estimated value, and then dividing the difference score by the GPS estimated value. All variables showed high ICC values (>0.75) and very large correlations (r > 0.70). However the video-based performance analysis system overestimated the results obtained in the different speed zones (DZ5: +16.59 ± 62.29 m; LOA95%: −105.49 to 138.68; DZ6: +93.26 ± 67.76 m; LOA95%: −39.55 to 226.07), the number of sprints (+2.27 ± 2.94; LOA95%: −3.49 to 8.02), and the maximum speed (+0.32 ± 1.25 km/h; LOA95%: −2.13 to 2.77). The maximum bias was found in DZ6 (47%). This demonstrates that Mediacoach® is as accurate as a GPS system to obtain objective data in real time, adapted to physical and movement demands of elite football, especially for total distance and distances traveled at medium speeds.
Real-time camera tracking using a particle filter combined with unscented Kalman filters
