Highlight events detection in soccer video using HCRF
Xueming Qian, Guizhong Liu, Zhe Wang, Zhi Li, Huan Wang · 2010
Highlight event detection is a fundamental step of semantic based video retrieval and personalized sports video browsing. In this paper, an effective hidden conditional random fields (HCRFs) based soccer video event detection method is proposed. Firstly, soccer video is classified into clips with middle level semantics. The middle level semantics are further refined into more meaningful categories in terms of camera motion information. Then the continuous soccer video sequence is classified into sequential event clips based on the transitions of middle level semantics. HCRFs are utilized to model the four highlight events (goal, shoot, foul, and placed kick) and a normal kick. Comparisons are made with the dynamic Bayesian networks (DBN) and conditional random fields (CRF) based event detection approaches. Experimental results show the effectiveness of the proposed method.