Remote Sensing Image Application Domain Knowledge Service Research Based On Knowledge Graph

With the development of earth observation technology and means,remote sensing images are increasingly showing a development trend of massive,multi-source,heterogeneous and distributed.As the extension application range of remote sensing images,different industries and users put forward higher requirements for the efficient and rapid service of massive remote sensing images.Facing the remote sensing images with abundant resources but lack of knowledge and the popularization of their applications,how to construct the domain knowledge semantic expression model for remote sensing image applications and provide personalized services,so as to meet people's multi-domain and multi-level needs for remote sensing images and knowledge,has become the difficulty of remote sensing image intelligent services.At present,remote sensing image services are mainly based on database technology and metadata,which lack the ability of semantic information description and understanding.Therefore,for the application of remote sensing images,knowledge graph related technologies are introduced to mine the relevant knowledge of remote sensing image application field and the logical relationship between various knowledge,and provide personalized services for various users relying on the knowledge.The main research contents of this article are as follows:1)According to the requirements and application characteristics of remote sensing images from different sources in different application fields,a knowledge system of remote sensing image application fields is built to effectively express the remote sensing image application knowledge from the aspects of space,time,remote sensing application tasks and remote sensing images.This paper uses the powerful semantic expression and reasoning capabilities of ontology to standardize the expression of concepts,attributes,and spatio-temporal semantic relationships between concepts in different application fields,eliminate semantic conflicts between multi-source heterogeneous remote sensing data,and promote the effective acquisition of remote sensing image information by users and the co construction and sharing of knowledge content;2)Based on the data model defined in the ontology of remote sensing image application domain,this paper has studyed the domain knowledge discovery process and related technologies for knowledge services.The main data source is the data resources(documents,encyclopedia data,industry website data,etc.)scattered in the Internet and the application cases of remote sensing images collected by industry departments,with the help of intelligence such as semantic analysis technology,neural networks,natural language processing framework semantic understanding,etc.Processing technology realizes the filling of knowledge units in the field of remote sensing image application,forming a small-scale domain knowledge base.3)A knowledge retrieval model of remote sensing image application field supporting semantic understanding and A personalized retrieval model of remote sensing image driven by knowledge are proposed.For the personalized service of remote sensing image application knowledge and image data,a semantic similarity measurement model for remote sensing image application field is proposed.In view of the lack of semantic understanding of traditional keyword search,with the help of the powerful semantic processing capabilities of knowledge graphs,semantic analysis and expansion of user input query statements are carried out to realize the efficient service of domain knowledge.Combined with the concept and thought of personalized information retrieval,information recommendation and other technologies,a remote sensing image combining retrieval" mode and "active push" mode is proposed4)In this paper,a user personalized modeling method for remote sensing image application is studied,and a spatiotemporal periodic model STPT is proposed,which can better express the user's information needs.Then the validity of the proposed model is proved by comparing with some existing models.Through the user's retrieval behavior records of remote sensing images,the joint distribution of potential application tasks in time,space and image characteristics is studied,personalized modeling for users in the field of remote sensing image applications is made.Users'retrieval behavior record of remote sensing images can be regarded as a combination of potential tasks and the connection between users and remote sensing images.Each application task is a joint distribution of space,time and image features.The joint distribution takes the Dirichlet distribution as a priori,and the von Mises distribution is introduced to express the distribution of application tasks over time.5)The knowledge retrieval system for remote sensing image applications based on knowledge maps is designed and implemented to realize the visualization of results and knowledge navigation.And then the personalized service system of remote sensing image data driven by knowledge is developed.The prototype system verifies the feasibility of the key technologies in the above chapters and achieves the expected results.