Sports fans and commentators in sports such as cricket, football and tennis 
have come to expect an increasing amount of real-time information about the 
game. Technology is also starting to play an important role in the umpiring of 
games, such a the video replays used in cricket or rugby, or the Hawk-Eye 
system now used in tennis for line calls. In the past, all statistics were 
gathered by hand and all decisions made by eye, but computer vision provides 
an ideal solution to the problem of doing this automatically. Importantly, 
computer vision can make use of existing infrastructure - the cameras around 
the ground - and is non-invasive: it requires no modification to the pitch, 
the equipment, or the players.

In this seminar, I will first give an introduction to some of the ways in 
which computer vision can be used in sports events. I will describe how the 
trajectory of a ball can be tracked in three-dimensions using two (or more) 
cameras around the ground, and also how the (two-dimensional) position of the 
players on the pitch can be determined using only one camera. The former 
theory is now routine, but I will then present new work that shows how further 
information can be revealed by using the shadows cast by the sun or by 
floodlights. These shadows have previously been considered a nuisance, to be 
removed before the real work can be performed. I will show how the casting of 
shadows can be modelled within the classic computer vision framework, and how 
each light source can be treated as being equivalent to a further camera. 
This means that with a single camera, full three-dimensional tracking can now 
be performed and, with a full set of shadows from floodlights, 
three-dimensional models of the players can be generated.