We present an algorithm for tracking the position of a curling stone from broadcast footage of curling events. Due to the constant panning of the camera in addition to perspective changes in the image, this introduces a need for image rectification and tracking of the camera view on the ice. We break the algorithm in four distinct components: rectification, frame tracking, stone tracking, and localization. The resulting algorithm has good performance in tracking the stone as it travels down the ice. This has promising applications for advanced analysis of curling shots, including the prediction of stone paths and precise quantification of the effect sweeping has on the stone.
© Copyright 2014 Published by Stanford University. All rights reserved.