Tracking soccer players using spatio-temporal context learning under multiple views

With the popularity of soccer games and rapid development of computer technology, automatic soccer analysis systems have been studied a lot these years. Tracking soccer players, as the fundamental step in an analysis system, is of great research value and draws attention from researchers all over the world. In this paper, we propose an effective method which makes an improvement on spatiotemporal context learning and increases the accuracy by combining information from multiple views. At the same time, a two-dimensional plane graph is displayed to show the players’ movements correspondingly. Experiments are conducted on several video fragments and the results have shown that the proposed method reaches a relatively high accuracy even when there are heavy occlusions and pose variations.