Grids – an Architecture for Managing Very Large Orthoimage Mosaics in a Database Framework

This paper presents the investigation and implementation of concepts for the efficient management of very large raster and orthoimage mosaics in a database framework. The paper first identifies and discusses the key requirements of a data management solution for orthoimage mosaics. These requirements include efficient support for spatial partitioning, spatial partition indexing and access, multi-resolution support, georeferencing, storage management and mass storage support. The paper then presents the architecture of the prototype spatial raster management system GrIdS. GrIdS is based on a 'spatial raster middleware' approach which allows typical GIS database engines, e.g. Oracle, to be extended with a spatial raster management capability. Key features of the architecture include a tile-based spatial partitioning solution and a spatial access and query concept using a modified Morton code-based indexing mechanism. The presented multi-resolution concept for raster mosaics represents a novel approach. It uses a virtual resolution pyramid of constant-dimension tiles, combining the resolution access with the spatial access through a single spatial coding scheme. The paper also presents a series of tests and results. These results demonstrate that orthoimages and very large orthoimage mosaics can efficiently be managed with state-of-art relational DBMS technology. The paper concludes with an outlook on important DBMS trends as well as on ongoing and future investigations.