Soccer Highlight Detection Based on Replay and Affection Arousal Model
Jun Yu · Chinese Journal of Computers · 2014
As one of the key technologies of content-based soccer video retrieval,highlight detection is of great academic value and broad application prospects.The existing replay and goal-mouth detection algorithms are improved in this work.The replay feature is also used to improve the event detection performance of affect arousal model.According to the needs of different users,a new highlight locating method is proposed for three different boundaries of highlight including short,medium and long boundary.The highlights of goal,shoot and foul could be recognized based on the game rules and the domain knowledge in soccer video.Experiments conducted on large amount of soccer videos show that the improved affection arousal curve which was fused with replay feature could reflect the fluctuations of the true soccer game process more exactly.The semantic event recognition especially the goal event recognition has higher recall and precision rate.