Real-time analysis scheme of football matches based on face recognition and multiple object tracking

The real-time analysis of football matches can bring more event information and wonderful visual experience to the audience. However, most of the existing game analysis is offline processing, and it is difficult to achieve real-time performance. This paper proposes a real-time analysis scheme for football matches based on face recognition and multiple object tracking. First, we classify the live broadcast data of the game, and then process the different scene data separately. Finally, the processed player game information is rendered into the video. For close-up images of game videos, we utilize face detection, face recognition, and optimized head tracking to determine player information. For distant shots, we propose a multi-object tracking algorithm based on player features to determine player associations in consecutive frames. This scheme has excellent performance in actual use. It can improve the user experience of watching games and is scalable.