Recently, several remote sensing IIM research prototype systems have being developed such
as Algorithm Development and Mining (Adam) System at ITSC [12], Diamond Eye System at
JPL/NASA [13], Intelligent Satellite Information Mining System at DLR [14], and VisiMine
System at Insightful Corporation [15]. Unlike Adam and Diamond Eye which are general data
mining systems for algorithms development and distributed processing in a wide range of
applications for earth science data, the integrated framework presented here is for experts or nonspecialists
to retrieve spectral and spatial information in remotely sensed image databases. The
key component in DLR system is a hierarchical naive Bayes learning model which lacks a
convenient query interface. We aim to build an adaptive system processing queries including the
remote sensing domain semantics from the user’s view point. VisiMine is the latest search engine
for analyzing image databases designed for satellite imagery and aerial photos. Its infrastructure
addresses the key scientific need for organizing and discovering information in large databases of
remotely sensed images. However, it stores features such as color, texture and shapes together
with raw images within the same relational database model. We discuss a different data modeling
approach, which stores features in an object-oriented database to facilitate the clustering and
retrieving, and stores the images in a separate image database to facilitate browsing and query