Using Neural Networks for Clustering on RSI Data and Related Spatial Data

Clustering is unsupervised classification of patterns into groups. Neural networks are very useful tools for performing clustering. In this paper, we propose a new model for using artificial neural networks to perform clustering tasks on remotely sensed imagery. This model generates self-organizing maps (SOM) based on remotely sensed imagery and such related data as yield, nitrate, and moisture. It correlates these maps and projects these outputs into a SOM. In addition, it uses wavelets for data pre-procession. The model also derives important rules and prunes unnecessary rules. The entire model is implemented as a distributed system using CORBA. Performance analysis shows the model is efficient and effective for performing clustering on remotely sensed imagery and correlated spatial data.