Comparison between the conformity coefficient and previous classification techniques for bare surface discrimination and application to compact polarimetry mode
My-Linh Truong-Loï(1), Anthony Freeman(2), Pascale Dubois-Fernandez(1) and Eric Pottier(3)
(1) ONERA, BA 701, 13661 SALON AIR CEDEX, France
(2) JPL, California Institute of Technology, Pasadena CA 91109, United States
(3) Université Rennes 1, 263 Avenue Général Leclerc, 35000 Rennes, France
In a previous paper, we have investigated the potential of the compact polarimetry mode at longer wavelengths in a space environment for soil moisture estimation. At longer wavelengths, one of the main challenges in recovering the information content in the measured scattering signatures, is Faraday rotation estimation and correction. In order to do that for bare surfaces using compact polarimetry scattering signatures, we have introduced a parameter, the conformity coefficient.
This criterion was observed to provide similar results to the Freeman-Durden classification in discriminating bare surfaces. In this paper, the study is pursued to include a more theoretical approach to the comparison. Furthermore, other published classifications as the one proposed by van Zyl, or the one based on the entropy-alpha-anisotropy are explored and the conformity coefficient is benchmarked against those other techniques. The comparison between these classification techniques is first performed theoretically and the effect of noise is investigated. The next step is then to apply all techniques to existing datasets coming from various sources, airborne sensors as RAMSES and AIRSAR and spaceborne sensors as PALSAR data.
The Faraday rotation angle can be estimated from full polarimetric data as was proposed by both Freeman and Meyer. From compact polarimetry data, this estimation is more difficult but we have shown in a previous paper how bare surface behavior can be used to perform this estimation. In this paper, the performance of this estimation is evaluated on PALSAR data and on airborne data where a constant ionosphere is simulated. The results are contrasted with estimates derived from full polarimetric data.