Meteorological classification of satellite imagery using neural network data fusion

A general approach to meteorological classification based on neural network data fusion is presented. The system implements a low-level vision system based on a number of biologically plausible theories operating across all input channels. Preprocessing stages derive products that are included with actual data as input to a classification stage. Supervised learning is used to train the classifiers. A number of promising preliminary results are presented, including a demonstration of robust classification performance over large shifts of sun angle and terrain. These results point to the applicability of neural networks for automated generation of meteorological products in real time