Indexing, Object Segmentation, and Event Detection in News and Sports Videos - Multimedia Database Retrieval: Technology and Applications - page 201

Fig. 7.23 A histogram of the number of shots within each time unit. Negative time units represent
early detection and positive time units represent a delayed detection

7.6.2 Classification of American Football Events

In the domain of American football, visual or motion features play a significantly
dominant role in discriminating between different types of plays as evident from
Fig. 7.21 . Therefore first we evaluate the efficacy of using motion descriptors for
an American football video indexing system and then we evaluate the changes in
system performance by adding audio descriptors and MFCC features.

7.6.2.1 Motion Feature Mapping

The motivation behind using the motion descriptors was due to the fact that in
American football, the global motion between different types of plays provides a
variety of clues. In order to understand fully the difference in motion between the
plays, first we require a detailed explanation of the general motion involved in the
plays:

1. Pass Plays : During a pass play, first the motion is lateral in order to track the
movements of a quarterback who is going to throw the ball. Then it is followed
by rapid zoom out and followed by a lateral movement to follow the throw.
At the end of the play the motion is tracking the player to whom the ball was
thrown. Therefore, the movements for a pass play involve first low intensity
lateral movement followed by high intensity zoom out and lateral movement and
then in the end low intensity lateral movement.