Event detection and semantic identification using Bayesian belief network

A probabilistic Bayesian belief network (BBN) based framework is proposed for semantic analysis and summarization of video using event detection. Our approach is customized for soccer but can be applied to other types of sports video sequences. We extract excitement clips from soccer sports video sequences that are comprised of multiple subclips corresponding to the events such as replay, field-view, goalkeeper, player, referee, spectator, players´ gathering. The events are detected and classified using a hierarchical classification scheme. The BBN based on observed events is used to assign semantic concept-labels, such as goals, saves, and card to each excitement clip. The collection of labeled excitement clips provide a video summary for highlight browsing, video skimming, indexing and retrieval. The proposed scheme offers a general approach to automatic tagging large scale multimedia content with rich semantics. Our tests using soccer video shows that the proposed semantic identification framework is more efficient.