Knowledge-Discounted Event Detection in Sports Video

摘要: Automatic events annotation is an essential requirement for constructing effective sports video summary. Researchers worldwide have actively been seeking the most robust and powerful solutions to detect classify key (or highlights) in different sports. Most of current widely used approaches employed rules that model typical pattern audiovisual features within particular sport events. These are mainly based on manual observation heuristic knowledge; therefore, machine learning can be as alternative. To bridge gap between two alternatives, we propose a hybrid approach, which integrates statistics into logical rule-based models during highlight detection. We also successfully pioneered use play-break segment universal scope detection standard set applied sports, including soccer, basketball, Australian football. The proposed method uses limited amount domain knowledge, making this less subjective more An experiment using large data has demonstrated effectiveness robustness algorithms.