A computer vision system to detect diving cases in soccer

Recently, motion analysis systems have been getting a lot of attention due to their potential in human motion analysis, which has a wide range of applications. One of these applications is analyzing tackle scenes in soccer games. In a tackle scene, players occasionally tend to deceive the referee by intentionally falling to get a free or penalty kick. In this paper, we propose a system to program human body tracking in order to analyze tackle scenes in soccer games. The main idea behind this system is to determine whether the falling player in the tackle scene is attempting to deceive the referee (diving) or not. In this system, the tackle scene goes through five main stages of processing; identification of the falling player, extraction of tracking points, motion tracking, features extraction and scene classification. The tracking component is implemented using Kanade-Lucas-Tomasi optical flow with the aid of pyramid levels and forward-backward error algorithm, while the classification is carried out using Weka software with Naive Bayes tree (NB tree) classifier. The proposed system is implemented and its performance is experimentally tested. The results show a potential to detect diving cases (deceiving in falling), with a classification accuracy of 84%.