Abstract
The ball trajectory carries a lot of information for tactics analysis in team sports, so we are interested in automatic ball tracking in American football video. While there arc many automatic ball tracking methods, because of holding the ball by one of players and heavy occlusions by the ball holder and other players, tracking ball is difficult in American Football video. Therefore, we calculate the play initial and terminal points using initial formation detection and the concentration degree of each player's direction. After that we connect the ball trajectory or the ball holder's between the play initial and terminal points. We decide which player is the actual ball holder using the team identification result, the distance from play terminal point, and the receiving approach degree by using the information of all players' movement.
Original language | English |
---|---|
Pages (from-to) | 91-98 |
Number of pages | 8 |
Journal | Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering |
Volume | 81 |
Issue number | 1 |
DOIs | |
Publication status | Published - 2015 Jan 1 |
Externally published | Yes |
Keywords
- AdaBoost
- Ball tracking
- Dense optical flow
- HOG
- Sports video analysis
ASJC Scopus subject areas
- Mechanical Engineering