When broadcasting sports events it is useful to be able to place virtual 3D annotations on the ground, to indicate things such as world record lines and distances. This requires the camera pose to be estimated in real time, so that the graphics can be rendered to match the camera view. Whilst camera calibration data can be obtained by using sensors on the camera mount and lens, such sensors can be impractical or expensive to install, and often the broadcaster only has access to the video feed itself. An image-based method of tracking the camera movement is thus the only practical approach in many situations. This paper reviews past work on image-based camera tracking, and presents the method we have developed. Our approach uses a method based on randomized trees for initial feature identification, and a KLT-based tracker to track features from frame to frame. Results of our system are presented on a selection of material representative of typical broadcast athletics, and the performance benefits of the approach we have taken are compared to a simple KLT-based tracker. This document was originally published in the proceedings of the IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB 2009) Bilbao, May 13-15 2009. It is copyright IEEE and is reproduced here with permission. The slides used in this presentation are included in the appendix.

Parent project