Shot classification for association football from TV broadcasting

Abstract

The paper presents the shot classification algorithm used to classify the scenes from football TV broadcasting, to extract only the far view scenes. The far-view scenes are important for the segmentation and tracking of the players during the game, required for devising an abstract model for the interpretation of the game. The system for developing such a platform is named ACE-Football, and is currently being developed at the PRIS-lab. Results tested using whole games of the latest FIFA's World Cup (FWC) are presented, and the classification algorithm validated using standard metrics, such as precision, recall and F1-score.