Neural processing of multispectral and multitemporal AVHRR data

In this work a mixed NDVI data processing method has been developed. It uses both statistical algorithm and neural network techniques to process and analyse the large historical archive of NDVI data being acquired by the NOAA/AVHRR sensor and developed by FAO-ARTEMIS jointly with NASA-GSFC. The archive contains ten years of data, so that it is possible to analyse, within this wide temporal range, the spatial and temporal variation of the vegetation index. The Principal Component Analysis has been used to reduce the amount of data to be processed. A neural network has been used to produce the clustering map. Some statistical parameters have been extracted from each cluster and the results have been compared with the ones obtained by statistical clustering (ISODATA algorithm).