Multi-level Indexing and GIS Enhanced Learning for Satellite Imageries

Satellite technology produces data at an enormous rate. Most of the database research on the analysis of remotely sensed images concentrated on data retrieval and simple queries that involved spatial joins and spatial selections. For example, the Sequoia 2000 project [13] aimed at the retrieval of raster data, while the Sloan Digital Sky Survey [14] poses the need for the creation of multiterabyte astronomy archive. The large scale systems for the analysis of remotely sensed images were specialized toward the detection of particular features like volcanoes [2], or proposed distributed and parallel data storage and query processing systems for handling of geo-scientific data retrieval queries [11]. The GeoBrowse project aims to provide infrastructure that would enable the analysis of large databases containing satellite images. Our work addresses two issues. One is the extraction of information that enables reduction of the data from multi-spectral images into a number of features. Second is the organization of the features that would allow flexible and scalable discovery of the knowledge from the databases of remotely sensed images. In this paper we present the concept of data mining system for the analysis of satellite images and preliminary results of the experiments with the collection of LANDSAT images.