The Research And Application Of Content-based Semantic Extraction Software For Videos

There exists the"semantic gap"between low-level features and high-level semantic because of the complexity of video semantic content. If we extracted the high-level semantic that consists with human requirements from the video and retrieve through using semantics to describe the characteristics of video information, we would found an effective way to solve"semantic gap"between low-level features and the high-level semantics. This paper mainly researched the extraction of the text information and audio feature aiming at"semantic gap", and then proposed the method of multi-feature fusion to achieve video scene segmentation according to the respective characteristics of the news video and sports video. We established soccer video domain ontology and adopted temporal logics to define timing relationships of sequence frames in football video, and then seted up inference rules based on the characteristics of semantic events of football video to achive detection and semantic annotation of football events. At last, this paper designed the video retrieval system that based on the semantic content.