Automatic classification of tennis video for high-level content-based retrieval
Lee, John Chung-Mong
Jain, Anil K.
|Summary||This report presents our techniques and results on automatic analysis of tennis videos for the purpose of high-level classification of tennis video segments to facilitate content-based re-trieval. Our approach is based on the generation of a image model, valid under perspective projection, for the tennis court-lines in a tennis video. We first derive this model by making use of knowledge about dimensions of a tennis court, typical camera geometry used when capturing tennis video, and the connectivity information about the court lines. Based on this model, we develop a court line detection algorithm and a robust player-tracking algorithm to track the tennis players over the image sequence. In order to select only those video segments which contain tennis court from an input raw footage of tenn is video, we also propose a color-based tennis court clip selection method. Automatically extracted tennis court lines and the players' location information form crucial "measurements" for the high-level reasoning module which analyzes the relative positions of the tennis players with respect to the court lines and the net, in order to map the measurements to high-level events (semantics) like "baseline-rallies", "passing-shots", "serve-and-volleying", "net-games", etc. Results on real tennis video data are presented demonstrating the validity and performance of the approach.|
Files in this item: