Discrimination of first year sea ice deformation features using synthetic aperture radar

Multiple image classifications are performed to examine the effect of system parameters (i.e. polarization and incidence angle) and image processing techniques (i.e. texture and adaptive filters) on discriminating first year sea ice deformation features in ENVISAT ASAR imagery. Results indicate that images acquired at higher incidence angles improve the discrimination of sea ice deformation features. Co-polarizations images: VV (khat=0.576) and HH (0.559) produce considerably higher classification accuracies than the cross-polarizations HV/VH (0.300). Applying speckle filters to the ENVISAT ASAR imagery is found to considerably improve classification accuracy. In an attempt to further improve classification results, texture measures are incorporated with speckle filtered imagery but lower the classification accuracy in all cases. These results indicate that the optimal choice for discriminating FYI sea ice deformation from 12.5m resolution ASAR data is gamma filtered (7x7 window size) HH VV polarization imagery acquired at swath 7 (∼45°).