#85: A novel framework for semantic annotation of soccer sports video sequences

M. H. Kolekar, K. Palaniappan, and S. Sengupta

Proc. 5th IEEE European Conference on Visual Media Production, pgs. 1-9, 2008

fmv, features, fusion, visual events, data mining, cbir

PlainText, Bibtex, PDF, URL, DOI, Google Scholar


A novel framework is presented for semantic labeling of video clips, automatically segmented from broadcast video of soccer (football) games, as highlights and excitement clips etc. The proposed framework provides a generalizable method for linking low-level video features with high- level semantic concepts defined in a commonly understood sports lexicon. Three important contributions are made to automatic annotation of sports video, as follows. First, domain knowledge combined with an event-lexicon and a four-level hierarchical classifier based on low-level video features is used to label video segments. Second, a priori event mining is used to establish probabilistic event-associations that are used to assign a concept-lexicon, such as goals and saves, to each highlight video segment. And, finally, the collection of highlight video clips is summarized using concept- and event- lexicons to facilitate highlight browsing, video skimming, indexing and retrieval.