Research On The Key Techniques Of Cloud-based Remote Sensing Data Management And Production Across Multi-data Centers

The rapid development of Earth observation technology,not only caused the explosive growth of remote sensing data,the kinds and quantities sharp increasing of remote sensing products,but also brought great challenges to the data storage and management,data processing and production systems.For remote sensing data integration and management system,massive,multi-source and heterogeneous remote sensing data not only brought difficulties for the unify and realization of metadata integration,but also made the existing remote sensing data storage and management system could not meet the demands of fast retrieval and high concurrent access.For data processing and production system,on one hand,the massive and multi-source remote sensing data easily makes the trouble of the most suitable data selection;on the other hand,due to the fact that no single remote sensor can cover such a large area at one time,there is often a lack of sufficient special data for a long-term series observation of a large area of Earth.Furthermore,existing remote sensing product services primarily support one or two remote sensing data sources,and these services are not capable of providing users with a variety of remote sensing data sources for selection.Therefore,using the advantages of large-scale data storage,high-performance cluster computing,elastic expansion and on-demand services of cloud computing,aiming at key problems of remote sensing data management,processing and production across multi-data centers,this paper carried out the following researches:(1)In view of the problem of remote sensing data integration and management,we first introduced the geometric information metadata standard to establish the distributed remote sensing metadata integration criterion,and then realized the unified format conversion of multi-source remote sensing metadata;Secondly,by establishing the spatial block division based index for multi-source remote sensing images,the spatial correlation and their corresponding metadata retrieval efficiency were improved;Thirdly,through the establishment of distributed index,hot metadata cache and high concurrent access control mechanism,the rapid retrieval of massive remote sensing metadata was achieved;Finally,through the establishment of virtual mapping between the public storage data and user owned data,we realized the multi-source remote sensing data sharing and personalized customization in cloud storage.(2)In order to solve the problems of remote sensing data processing and production,we first established a data recommendation model for multi-source remote sensing product generation,so as to provide theoretical basis for multi-source remote sensing data optimization selection during the process of product generation;Secondly,we built the upper and lower hierarchical relationship database,the product dependence database,et.al,and constructed the logical process of product generation;Finally,we constructed a cloud-based remote sensing production environment to provide massive remote sensing data processing and production platform,in order to improve efficiency of the mass remote sensing data processing and production.The significances of this paper are: First of all,the unified metadata format conversion,shielded the differences of heterogeneous remote sensing metadata in distributed data centers,and laid a foundation for the integration;Secondly,by studying the spatial organization pattern,the remote sensing data identifier is assigned by geo-correspondence function,and the spatial correlation between multi-source remote sensing data is also improved;Thirdly,the establishment and optimization of distributed index,improved the efficiency of multi-source and large-scale remote sensing data retrieval and distribution services;Moreover,the research of multi-source remote sensing data virtual mapping and sharing,not only satisfied the individual needs of different users,but also maintained the consistency of cloud data storage;Finally,the development of multi-source remote sensing data recommendation,the logical production workflow organization and the multi-source remote sensing production across multi-data centers,could improve the scientific nature of remote sensing information product services and explore the feasibility technology for remote sensing information service.