Goal Event Detection in Soccer Videos Using Multi-Clues Detection Rules

This paper proposed a framework for goal event detection in soccer videos by using multi-clues detection rules. In this framework, the visual clues including shot segmentation, shot classification and goal detection is extracted. Meanwhile, the audio clues including the audience´s cheering and the commentator´ excited speech are extracted. Then the goal event detection rules are defined by combining the visual clues, audio clues, the program composition and editing rules as well as the domain knowledge of soccer video. Experiments on real soccer video programs demonstrate encouraging results.