基于本体的多模式融合语义提取模型

摘要

There're two key technologies for semantic based video retrieval has to deal with. One is to bridge the semantic gap between low-level features and high-level semantic concepts. The other is to build a effective semantic model. In this paper,based on the multi-level semantic analysis of video,employ an effective object segmentation method on extracting the semantic object in video,and use se-mantic object as the middle level and consider the fusion of multi-model,such as image,sound and text,to bridge semantic gap. Mean-while,videos contain multi-granularity semantic concepts. And since it has superiority to describe concepts and relationships between them,propose a ontology-based semantic model,to reason compound concept of higher level from atomic concepts,which are extracted from image,sound and text. By using this semantic extraction model,can get the video semantics contain richer semantic level and se-mantic granularity. Thus it can be closer to the semantic concepts in human thoughts.%基于语义的视频检索要处理的两项关键技术就是解决视频低层特征和高层语义概念间的语义鸿沟以及有效的语义提取模型。文中通过对视频进行多层次语义分析，采用有效的语义对象分割方法提取视频中的语义对象，以语义对象为中间层，并融合图像、声音、文本的多模式视频特征，从而缩小语义鸿沟。其次，视频语义概念具有多粒度性，由于本体在表示概念及概念间联系时的优越性，文中提出基于本体的语义提取模型，在从图像、声音、文本中提取出的原子概念中，推理出更高层次的复合概念。最终运用该模型提取的视频语义就具有更丰富的语义层次和语义粒度，从而更接近人类思维中的高层语义概念。