Artificial vision and augmented reality applied to the analysis of sports broadcast videos

Tracking players and identifying them from video sequences of sport events, taken from a unique Pan-Tilt-Zoom (PTZ) camera, are exploited and experimented in various applications. However, this challenge is still difficult to realize and presents a vast field to investigate. In this paper, we suggest an automatic system that tackles this quiet complex problematic. The system has the ability to detect and follow multiple actors, estimate the homography between successive images of a video sequence and the play-ground and identifying the actors in the scene. The homographic estimation is solved by using a variant of ICP algorithm (Iterative Closest Point). The results of these applied artificial vision operations in the field of sport analysis open tremendous possibilities of applications and of mixed and increased reality in the benefit of sport events broadcasted on television.