<![CDATA[This research develops a computer vision‑based football player speed analysis system integrating YOLOv8 for object detection, DeepSORT for multi‑object tracking, and K‑Means clustering for team identification based on jersey color in the HSV color space. The system processes match videos through preprocessing stages (resizing, color space conversion, and filtering), camera motion compensation using Lucas‑Kanade Optical Flow, and speed calculation with filtering and smoothing mechanisms. Testing on a 30‑second match video demonstrates that HSV‑based K‑Means clustering effectively distinguishes players into two teams with a Silhouette Score of 0.62 and a centroid distance of 5.71°, despite highly similar jersey colors. Speed estimation yields average speeds of 19.7–23.47 km/h, which fall within the high‑speed running category (19.8–25.1 km/h) based on literature, with a maximum speed of 36.2 km/h aligning with professional sprint performance. Comparative analysis proves that HSV is more robust to lighting changes compared to RGB, which fluctuates up to 23 points. The system is implemented as a Flask‑based web application enabling video upload, automated analysis, and annotated video download. Therefore, the system is feasible as an objective and measurable tool for player performance analysis. ]]>
Copyrights © 2026
