Video-assisted refereeing technology (VAR) has been widely used in various large football events. The introduction of this technology can effectively reduce the misjudgment rate and ensure fairness. However, the essence of this technology is to use video slow action playback. For foul judgment, human eyes are still used. In addition, the use of this technology will interrupt the process of the game and reduce the net time of about 4 % of the game, which will greatly reduce the user's viewing experience. In this paper, the target detection and pose extraction technology of artificial intelligence in recent years are integrated into the traditional VAR, and the migration learning of the YOLO network is carried out to realize real-time detection. The specific parameters of handball detection are as follows: recall 0.840, precision 0.913, and detection speed 8.2 frames/second.
© Copyright 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology. All rights reserved.