Semantic Units Based Events Detection in Soccer Videos (TA-P6)

Author(s) : Xiaofeng Tong (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, China) Qingshan Liu (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, China) Hanqing Lu (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, China)

Abstract : In this paper, a three-layer event detection scheme based on semantic units is proposed and applied to shoot and card events detection in soccer videos. At the lowest layer, low-level features including color, texture, edge, motion and shot boundaries are considered. The highest layer involves semantic events. In order to connect the low-level features and video semantics and bridge the semantic gap, we present a concept of semantic unit and define several semantic units according to domain-dependent knowledge and specific events at the middle layer. Semantic unit is consecutive frames tagged a special semantic cue. Taking the semantic units as observations, Bayesian inference based probabilistic framework is used to reason the probabilities of interested events. The experimental results demonstrate that the scheme is effective and the semantic units extraction is reliable.

Menu