Soccer field detection in video images using color and spatial coherence

Abstract

We present an original approach based on the joint use of color and spatial coherence to automatically detect the soccer field in video sequences. We assume that the corresponding area is significant enough for that. This assumption is verified when the camera is oriented toward the field and does not focus on a given element of the scene such as a player or the ball. We do not have any assumption on the color of the field. We use this approach to automatically validate the image area in which the relevant scene elements are. This is a part of the SIMULFOOT project whose objective is the 3D reconstruction of the scene (players, referees, ball) and its animation as a support for cognitive studies and strategy analysis. © Springer-Verlag 2004.