Classification strategies for polarimetric SAR sea ice data

• Main Authors: Bernd Scheuchl, Irena Hajnsek, Ian Cumming
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Subjects: Sea ice
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.407.1931; http://sar.ece.ubc.ca/papers/PolInSAR_strategies.pdf

ABSTRACT / RESUME Bayesian classification of polarimetric data is based on the complex Wishart distribution of the coherency matrix. The classifier has shown to be a flexible tool available in a variety of options. In this paper we investigate several classification strategies for sea ice classification. All methods show reasonable results, although some user input prior to classification seems beneficial. Noise and incidence angle were found to affect the classification result, the latter due to a change in the relative mix of scattering mechanisms. Surface scattering is found to be strongly incidence angle dependent. Multi-frequency data have a superior information content and can separate ice types, whereas the scattering information content can not be exactly discriminated using only C-band data. 1