Real-Time Football Analysis System using YOLO and OpenCV

AI and ML technologies have increasingly dominated sports analytics, allowing new real-time data processing capabilities. This paper proposes a new Real-Time Football analysis system using YOLOv5 and OpenCV to respond to issues revolving round accurate Player and Ball detection with occlusion and illumination changes. To accomplish the methodology, object detection is performed, SORT algorithm for tracking and Kalman filtering for improving the tracking while compensating the camera movement using optical flow. The results demonstrate high accuracy in player and ball detection (precision: 92. It outperforms other methods (precision: 5%, recall: 89. 8%, mAP: 91. 1%) while implementing a real-time analysis at 28.5 Frames Per Second. Speed, distance covered, team possession, and more, are displayed, making the system excellent for detailed tactical analysis, potential improvement of individual player or team performances, and even rising enthusiasts’ engagement. In conclusion, this work presents a high-quality real time eight-layer analysis solution for football that can open the way to higher level predictive models in the future.