Research On Distributed Management Technology Of Global Mass Remote Sensing Image Data

The air and space remote-sensing (RS) image data of global area is a massive, multi-dimensional, multi-sources, multi-temporal data set. With the fast development of RS technology, the scale of the data set is growing in a surprisingly high rate. It makes the efficient and effective management of global distributed multi-dimensional RS image data more and more difficult. At the same time, in the application domain the demand of RS image is also growing quickly. The application should be capable of providing users with the image in the shortest time according to their query conditions including time, geo-position, spectrum band, quality, scale, format and size. But until now, this kind of requirements can not be met yet. The key bottleneck is the management technology of global RS image data is developing slower than the growth rates of RS image data volume and the demand of it. The intention of this subject is to mine the key technologies of global multi-dimensional mass RS image data management, and try to find the solutions.To be a system that is capable of managing global multi-dimensional RS image data, it should have four key technical features: Massive, Multi-dimension, Integrated capability and Temporal variability (MMIT). From this point of view, taking into account the features of RS image data and the applications, the technical framework of such an RS image data management system can be sketched. And the GARIMS is an implementation of this framework.The term dissection of global RS image data refers to two aspects: the multi-resolutions organization and the data partition on several dimensions of the RS image. The dissection includes two levels. The first one is operated mainly on temporal dimension along with the type of sensor. In this level of dissection, uniform and non-uniform partition strategies are researched. Temporal multi-resolution organization is also explored. The sub RS image data sets generated by the first dissection are called RIDAP. The second level of dissection is inside the RIDAP, aims at spatial dimension. It includes MPPQT, PCPG and PCPGQT proposed in this paper. The GLEDAM is a data model based on the two levels of dissection and is able to describe many other geospatial data types. It is good in scalability because its description of the geospatial data is on a higher abstraction level.The storage tier consists of a cluster built from many commodity components and a distributed file system built upon the cluster. This architecture is suitable for the storage of RS image data. The inner structure of RIDAP is independently designed, so that it can take advantage of the distributed file system. It has been proved that to manage the global RS image data without an existing RDBMS can bring several advantages. To solve the problems of seamless mosaic and data fragments collection in the process of building and maintaining the RIDAP, some creative solutions have been proposed.The methods of dissection organization, the solutions of building the RIDAP and the processing of image tile query are all implemented in the class library Smartlmagery. The experiments have proved the feasibility reliability and high performance of the methods and solutions.