Ron Caves
In anticipation of fully polarimetric SAR data from RADARSAT-2, we use a SIR-C data set to investigate the potential of fully polarimetric spaceborne data for sea ice classification. This paper discusses an entropy/anisotropy/α−angle (H/A/α) classification scheme followed by a minimum-distance class...
Main Authors: | , , , |
---|---|
Other Authors: | |
Format: | Text |
Language: | English |
Subjects: | |
Online Access: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.406.9340 http://sar.ece.ubc.ca/papers/CSRS_iceclass.pdf |
Summary: | In anticipation of fully polarimetric SAR data from RADARSAT-2, we use a SIR-C data set to investigate the potential of fully polarimetric spaceborne data for sea ice classification. This paper discusses an entropy/anisotropy/α−angle (H/A/α) classification scheme followed by a minimum-distance classifier based on the complex Wishart distribution of the coherency matrix. Fully polarimetric data acquired by SIR-C over the Labrador Sea off Newfoundland's West Coast is analysed. Data from the two available frequency bands, L- and C-band, are classified separately and the results are compared. Both classifications provide sea ice – open water discrimination and sub-classification of various sea ice types. The main differences between the two results are the number of sea ice classes derived (four for C-band, three for L-band) and the fact that L-band data seems to provide a little better ice/water discrimination. High correlation between the two results with respect to the main feature types (sea ice, water and land) can be noted. Although ground truth is not available, the results are considered of good quality because of the agreement between the L- and C- band results, and consistency with expert human interpretation. |
---|