Effective spatio-temporal analysis of remote sensing data

Extracting knowledge and features from a large amount of remote sensing images has become highly required recent years. Spatio-temporal data mining techniques are studied to discover knowledge from these images in order to provide more precise weather prediction. Two learning granularities have been proposed for inductive learning from spatial data: one is spatial object granularity and the other is pixel granularity. In this paper, we propose a pixel granularity based framework to extract useful knowledge from the remote sensing image database by using SOM and association rules mining. A three-stage algorithm, named as STARSI, is also proposed and used in this framework.