Approach for Data Acquisition Based on the Homogeneity of the Playing Field

Abstract

Nowadays, tracking and identifying actors in many areas and especially in team sports video has become a real challenge. In this paper, we propose a distributed real-time solution that tackles this problematic. This solution has the ability to supports the probabilistic fusion of different information sources, structures the incoming perceptions by automatically building models of the tracked players and adapting these concepts online. Regarding the identification of players we used adaptive methods based on positions and appearance, a method for summarizing team behavior from spatio-temporal data. This opens tremendous possibilities of applications of mixed and augmented reality for sport events and soccer games on TV at real time.