Offense based temporal segmentation for event detection in soccer video

  • Authors:

  • Lei Wang;Michael Lew;Guangyou Xu

  • Affiliations:

  • Tsinghua University, Beijing, P.R.China;Leiden University, Leiden, Netherlands;Tsinghua University, Beijing, P.R.China

  • Venue:
  • Proceedings of the 6th ACM SIGMM international workshop on Multimedia information retrieval
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Sports video is regarded as a good testing bed for techniques on content based video analysis and processing. Although partially successful systems have been designed for specific sports domains with limited data, most previous works do not adequately address the problem of temporal segmentation for event detection, nor the event representation problem. In this paper, we present an analysis of soccer video for detecting the semantic notion of offense. It is not only useful as a new semantic concept of sports video analysis, but also provides temporal segmentation for video event detection and representation. We propose a system to detect the offensive unit in soccer video automatically. The offensive unit is then used to calculate new semantics like possession, as well as to detect goal events in video. Experimental results on various sources of soccer video have verified that our approach extracts the new semantic notions successfully and facilitates video event detection and representation