利用本体的视频语义概念检测

Semantic Information Extraction of Video Using Ontology

视频数据的不断丰富以及人们对视频检索的要求越来越复杂,使得视频语义信息建模和高层语义概念提取逐渐成为视频检索中的重要组成部分.本文提出一种基于本体的视频语义概念检测方法,利用贝叶斯网络构造视频中概念语义关系的检测本体,构建了视频中概念之间的层次关系,并能够通过推理完成复合语义概念的检测.该方法从语义信息学的角度对视频内容进行分析,在一定程度上削弱了语义鸿沟的影响,并且取得了较好的查询结果.

With the growing of video data volume, the demand for video retrieval and video semantic information extraction grows as well. In this paper, a new ontology-based composite concept extraction method is proposed, which adopts Bayesian network to construct the ontology and uses the inference rules to perform the composite concept detection providing the atomic concepts in a segment of video. This method analyzes the video content through semantics and constructs ontology to express the hierarchical relationship between the concepts, which narrows the ＂Semantic Gap＂ and achieves good performance in abstract concepts extraction.