Research On Video Semantic Representation Method Based On Graph Structure From The Perspective Of Evolution

In recent years,the development of mobile Internet technology has greatly facilitated the production and dissemination of video content.Video content is affecting the information life of a large number of people in China,and the research and application requirements of video information resources are becoming more and more in-depth and detailed.The in-depth research and exploration of the value of video content must involve the semantic information of video.The latest research in the fields of video resource management,video data mining,video retrieval and so on cannot do without the attention to video semantics.The key to the above researches and applications is a suitable method of video semantic representation.However,there are some deficiencies in the existing video representation methods.Their semantic abstraction mechanism and temporal information representation capability need to be improved.They also lack the semantic extension mechanism from the evolutionary perspective and the general formal description model.So they cannot effectively represent the semantics of video,which is not conducive to the further researches and applications of video information resources.Therefore,this paper proposed GSVideo(standing for Graph Semantic of Video),a video semantic representation method based on graph structure from the perspective of evolution,in an attempt to explore solutions to the existing problems.In this article,we first clarified and analyzed the solutions to the the key issues in video semantic representation with an example video.The key issues are the bottomup evolution process,video semantic expansion from the perspective of evolution,topdown semantic completion,and user-oriented video semantic retrieval.On this basis,this paper put forward the video semantic representation method GSVideo.We defined the semantic units,semantic connections and formal description.We also gived the data operation rules such as data modeling and data query.Then we represented the semantics of the example video.After that we interpreted the semantic representation process combining with the mentioned key issues and futher discussed the detailed mechanism of the method.GSVideo is a video semantic representation method proposed from the perspective of evolution,which is based on the modeling primitive of graph model and the abstract mechanism of semantic data model.This method integrated high and low level information of video content and adopted semantic units with high cohesion and low coupling.We designed various semantic extension mechanisms for semantic evolution and also made time information fully expressed.In Addition,we provides an effective video semantic retrieval scheme general formal description method.In this paper,the GSVideo method is used to represent the semantic of the sample video.The method showed strong feasibility in semantic representation and semantic retrieval,which proved the effectiveness of the method.This paper also compared GSVideo method with the existing methods in the aspects of high-level and low-level information integration,semantic abstraction mechanism,semantic evolution,multi granularity time representation,formal description and so on.GSVideo method has obvious advantages in comparison,which explains the innovation of the method.