Scene Representation And Retrieval Of Security Video Content Based On Knowledge Element

With the rapid development of intelligent video analysis technology(IVA),intelligent security video surveillance system can effectively parse and extract more features and semantic information about video content.These extracted security video content information can help the users to carry out video content retrieval,linkage tracking,intelligent monitoring and other follow-up security applications.However,by different intelligent security video surveillance systems,the extracted video content information is expressed in different description forms,and also contains rich and complex event information,so that this leads to inefficient use of security video content information.Therefore,how to represent and store the security video content efficiently and also design an appropriate retrieval method is the needed scientific research problem to improve the security of video content retrieval efficiency and application value.Security video content information is the knowledge source that help users to do security prevention and control,decision-making.Its standardization organization and management will facilitate the rapid and accurate content retrieval and analysis,so as to meet the users' needs that using the video content information efficiently.This paper presents a scene representation and retrieval method of the security video content based on knowledge element.Firstly,from the perspective of common knowledge management,this paper analyzes numerous security video content information identified by intelligent video analysis technology,combs and summarizes the common knowledge,forms the knowledge element model of security video content,and builds the extensible security video content management knowledge element system to achieve the unified representation and integration application for the knowledge of the security video content that involved in the multi-disciplinary,multi-domain;Secondly,from the visual recording of the objective world to the subjective cognition of human beings,three-space mapping model is constructed,then according to the event transfer in security video content information,the security video content is divided based on the granularity of the scene,and obtains the scene fragments.And then combining the features of security video event shot detection,the security video scene is divided into foreground,background and basic information,combining with the knowledge element system for structured representation and designing XML format for its structured storage,security video content is expressed by the security video content information description library;Thirdly,based on security video content scene organization proposes a security video content retrieval method driven by time-space characteristics and then designs the corresponding retrieval process and similarity calculation method to achieve the accurate and efficient retrieval of the security video content scene.Finally,through the collection of a lot of real security video content knowledge elements to build a unified knowledge representation system,develops the social security video comprehensive analysis and early warning system to verify the feasibility and effectiveness of the methods proposed by this paper.In this paper,the scene representation and retrieval methods of security video content based on knowledge element are proposed,through constructing the unified knowledge element system of security video content information,effectively solves the complexity of the description of security video content.According to the division of the scene granularity and combining with a unified knowledge representation system for structured representation and XML form of storage,and then designs the retrieval method which conforms to the time-space characteristics of the video content.It provides strong support and guarantee for accurate and quick retrieval and analysis of the information needed.These methods greatly enhance the use efficiency and application value of security video content information.