Multi-Band Supervised Classification for Polarimetric SAR

Abstract

This work addresses the potential of multi-band polarimetric SAR imaging for terrains and vegetation classification. A classic supervised Wishart classifier is adapted to polarimetric multi-band datasets, and is applied on the X-, L- and UHF-band acquisitions done during the NAOMI campaign (ONERA-Total) in Gabon (Africa) in 2015. The contributions of the different frequencies are shown and discussed. It is shown that the use of the multi-band dataset improves significantly the classification result.