Abstract

Hybrid-polarity SAR architecture will be included in future SAR missions such as the Canadian RADARSAT Constellation Mission (RCM). These sensors will enable the use of compact polarimetry (CP) data in wide swath imagery. We investigate the sea ice type and open water classification potential of CP parameters for future RCM image modes. We simulate twenty three CP parameters for three RCM modes with different resolutions and noise floors, using RADARSAT-2 quad-polarized data. A case study is used encompassing multi-year sea ice, first-year sea ice and open water, in the Canadian Arctic. We use a methodology for quantitatively selecting the optimum CP parameter subset for classification. Preliminary CP parameter selection is based on Jeffries–Matusita separability. A batch-mode Maximum Likelihood classification approach is then used to iterate through all possible combinations of CP parameter subsets, in order to select the optimum subset, the one with the highest accuracy. Combinations between two and ten CP parameters, depending on RCM mode, are found to provide the highest accuracy. The effects of an increased noise level and higher resolution on different CP parameters are found to vary. However, the impact on classification accuracy is small. High classification accuracies support methodology for the selection of the working subset of the CP parameters and their classification potential. Multi-year sea ice, first-year sea ice and open water can be successfully classified with an overall classification accuracy of 99.92% using two simulated compact polarimetric parameters from the RCM LN25 mode: the circular cross polarization ratio and the volume scattering parameters of the m-χ and m-δ decompositions. In the same mode, the ice and the open water can be successfully discriminated with an overall accuracy of 100% using the sigma nought backscattering coefficients σRH∘ and σRR∘ in addition to the two previous compact polarimetric parameters.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call