Research On Large-Scale Distributed Remote Sensing Data Management Methods

With the development of earth observation technology and remote sensing technology,a remote sensing application system integrating heaven and earth has been initially formed in China.Different remote sensing data application research institutions have accumulated a large amount of remote sensing resource data and stored in their data centers.However,due to the different application fields of different data centers,their model organization,system architecture and management mode are also different.Such decentralized storage mode according to industry eventually leads to the single application form of resource data,which greatly restricts the sharing efficiency and application efficiency of remote sensing data.In view of the above problems,the integration of resource data from multiple data centers and the efficient sharing of data is the premise of data mass application.The federated service sharing across organizations and data centers is a low cost and efficient way of sharing.Therefore,this paper presents a virtual distributed management method for large-scale remote sensing data.Firstly,a unified and heterogeneous remote sensing metadata model is proposed.In this paper,a unified remote sensing metadata model is constructed by combining the concept of UMM remote sensing metadata mapping standard and the remote sensing metadata standard ISO19115-2,and then a global data resource directory is constructed on this basis.Secondly,based on the analysis of different retrieval behaviors of users,the corresponding retrieval strategies were proposed for different retrieval backgrounds,and the ir-tree hybrid index structure with text first was optimized by FST Tree,and its effectiveness was verified by comparative experiments.Finally,in order to achieve different satellite data center mass remote sensing data integration,this article is based on Alluxio distributed remote sensing data to build a virtual storage system,and combine with the idea of remote sensing metadata attributes to build a unified image data organization model,through the SVM to receive a new remote sensing data journal captures images and user behavior analysis,prediction of new hot spot image data and combined with Alluxio allocation strategy and recycling strategies will focus in the data cache,this strategy can effectively increase the user access to hot data.On the basis of the above technical research,the large-scale remote sensing data management platform "FSDCloud" was designed,the overall design and detailed design were carried out,and the prototype of the platform was realized by using relevant technologies.The large-scale remote sensing data management platform has application value.