Ball localization and tracking in a highly dynamic table soccer environment

This article presents the development of a ball localization and tracking algorithm, that is to be applied in a highly dynamic table soccer environment. The described approach is based on an earlier survey paper on object tracking, where a general selection procedure on object detection and tracking techniques was proposed. Although the survey paper presents a variety of state estimation techniques for tracking, this article describes why these are not well suited for our specific application. For this reason, an IMM estimation technique is adopted that has not been applied in this highly dynamic context before. To evaluate the IMM estimator, it is compared to the well-known and commonly used Kalman filter, that has been optimally tuned for this specific application.