Semantic units based events detection in soccer videos

In this paper, an event detection scheme based on semantic units is proposed. The scheme can be characterized as a three-layer framework. At the lowest layer, low-level features including color, texture, edge, shape, motion, etc. are considered. High-level events are modeled and characterized at the highest layer of this framework. In order to bridge the semantic gap between low-level features and high-level semantics, we define some semantic units at the intermediate layer. A semantic unit, which is composed of consecutive frames with similar cue, is a description for video segment. It is deduced by frame-wise low-level features and taken as evidence of event reasoning. In event inference, a Bayesian network based on probabilistic framework is utilized. Experiments have demonstrated that the scheme is effective and valid.