A New Strategy Of Storage&Retrieval For Massive Tile Data Of Remote Sensing Images

Remote sensing images are regarded as the basic geographical information data inthe research plan of digital globe, the high efficiency of storage and retrieval is animportant method of the extraction and sharing of remote sensing data. Recently, withthe rapid development of high-resolution earth observation technology, the amount ofremote sensing images is dramatically increasing. As a result, the problem of how toeffectively organize, store and manage the massive remote sensing data, as well as howto achieve the rapid sharing and release has become the urgent issue in the area ofspatial information science.It is easy to find that using multi-resolution pyramid and blocking is an effectiveapproach for the real-time display and high-speed service by reviewing the research inthis area both home and abroad. However, multi-resolution pyramid will definitely leadto a large increase in redundant data because this approach is performed by sacrificingthe space in return for response time, which will multiply the image data.According to this, the effective storage and management for large-scale remotesensing tile data has significant meanings. By analyzing some kinds of famous systemsabout storage and management of massive remote sensing images and combining thespatial characters of tile data, this paper proposed a new strategy for storage andretrieval, which was based on lightweight file databases (embedded databases) and filesystem. This strategy utilized distributed file system to store the tile data, and in thesame time, the metadata of tile data was stored in the file databases, which was stored intree structure of the storage nodes. This is, the tile data and the metadata of tile was bothstored by distributed method so that the limit of centralized metadata retrieval can beavoided. The specific strategy of storage drew on the theory of Hash mod, but there wasa limit in security and availability of a distributed system based on Hash. Based on this,this research designed an improved distributed strategy called ‘Hash major-minorallocated number’. In addition, the reliability, expandability and load balance of thedistributed system using this strategy were qualitatively analyzed. Apart from that, therewere two different retrieval methods based on this storage strategy, one was precise search based on ‘directly addressing’ and the other was fuzzy search based on databasefiles and multi-computer parallel technology. This paper illustrated the retrievalprinciple, the method and the core programme. In addition a large number ofexperiments were implemented. The comparison of fuzzy retrieval experiments showedthat this new strategy based on database files and multi-computer parallel technologyhad much higher retrieval efficiency than that based on traditional commerciallarge-scale database retrieval method as well as met the retrieval request formagnanimity data source.Finally, we concluded the main contributions and innovation points of this researchand proposed the working plan of next.