Multi-Camera Analysis of Soccer Sequences

The automatic detection of meaningful phases in a soccer game depends on the accurate localization of players and the ball at each moment. However, the automatic analysis of soccer sequences is a challenging task due to the presence of fast moving multiple objects. For this purpose, we present a multi-camera analysis system that yields the position of the ball and players on a common ground plane. The detection in each camera is based on a code-book algorithm and different features are used to classify the detected blobs. The detection results of each camera are transformed using homography to a virtual top-view of the playing field. Within this virtual top-view we merge trajectory information of the different cameras allowing to refine the found positions. In this paper, we evaluate the system on a public SOCCER dataset and end with a discussion of possible improvements of the dataset.