Nowadays increasing attention has been paid to reasonable organization and effective management of vast amounts of
remotely sensed data for the goal of quick browse, convenient query and Retrieval-on-Demand service.
In this paper, in order to reach compromise among precision, efficiency and storage and to realize ROI coding, data
partition based on Nona-tree data structure and data compression based on JPEG2000 are adopted to organize and
manage original remotely sensed images. Afterwards, a prototype system in three-tier B/S mode is developed to test the
validity of our data organization and management strategy for content-based retrieval mentioned above. In this system,
texture-based and shape-based feature extraction algorithms based on wavelet transformation, math morphology and
other relative theory are applied. Corresponding feature descriptor and similarity calculation are also given. At last,
experimental results are given to show that the strategy proposed in this paper is valid, followed by brief conclusions and
future directions. The work of this paper is useful to push the development of geo-spatial information services and
promote content-based retrieval of remotely sensed images from experimentation to practicality.
|