Online Multi-player Tracking in Monocular Soccer Videos

Abstract

The tracking of players in monocular soccer videos is a challenging task because of numerous difficulties that can occur especially in TV broadcasts, such as camera motions, severe occlusion of players, or inhomogeneous lightning conditions. We propose a new robust method for multi-player tracking, which is based on finding local maxima on a confidence map. This map represents an ensemble of visual evidences, such as colors of the team outfits, responses of a HOG human detector, and grass regions in images. This combination of features allows for a robust online tracking procedure that does not require any further information about the camera calibration or other user input. In the evaluation using four representative datasets, our algorithm shows remarkable accuracy and outperforms a state-of-the-art pedestrian tracker.