Authors: Adrià Arbués-Sangüesa, Adrián Martín, Javier Fernández, Coloma Ballester, Gloria Haro
Publication: International Conference on Image Processing (ICIP)
Publication year: 2020
Given a monocular video of a soccer match, this pa-per presents a computational model to estimate the mostfeasible pass at any given time. The method leverages of-fensive player’s orientation (plus their location) and op-ponents’ spatial configuration to compute the feasibility ofpass events within players of the same team. Orientationdata is gathered from body pose estimations that are prop-erly projected onto the 2D game field; moreover, a geomet-rical solution is provided, through the definition of a feasi-bility measure, to determine which players are better ori-ented towards each other. Once analyzed more than 6000pass events, results show that, by including orientation asa feasibility measure, a robust computational model canbe built, reaching more than 0.7 Top-3 accuracy. Finally,the combination of the orientation feasibility measure withthe recently introduced Expected Possession Value metric isstudied; promising results are obtained, thus showing thatexisting models can be refined by using orientation as a keyfeature. These models could help both coaches and analyststo have a better understanding of the game and to improvethe players’ decision-making process.