Assessing the benefit of SAR Polarimetry for Land Cover Classification
Room 1058, A8 Building,
Farnborough, Hants. GU14 0LX,
(2) SarVision, KvK Arnhem 09112525, Wageningen, Netherlands
In this paper we demonstrate and validate the benefit of polarimetry for land cover classification. The performance of a number of supervised and unsupervised algorithms are presented. Two data sources are used: DLR -ESAR P and L-Band data of Alling, Germany and JPL AIRSAR P, L and C Band data of Flevoland, The Netherlands. Both datasets have comprehensive ground truth and this is used as validation. In addition to full polarimetry, the performance of L and C Band partial polarimetry (HH-HV and HH-VV combinations) is also assessed. Confusion matrices are presented for each algorithm and conclusions about the best frequency and polarisation combinations to use are drawn.