Coherent and Incoherent Scattering Decomposition in Terms of Roll Invariant Target Parameters : Applications to Wetland Classification Using C-Band SAR
Ridha Touzi(1) and Alice Deschamps(1)
(1) Canada Centre for Remote Sensing, 588 Booth St.,, K1A 0Y7 Ottawa, Canada
This paper reports on the use of polarimetric target scattering decomposition for wetland characterization. Cloude-Pottier’s incoherent target scattering decomposition  has been for the last decade the most used method for target scattering classification. Recently, the Touzi decomposition is introduced [3, 4] for coherent and incoherent decomposition of target scattering in terms of unique and roll invariant parameters. In contrast to the Cloude-Pottier decomposition, which uses a real entity, the so called Cloude alpha, to describe target scattering type, the Touzi decomposition characterizes uniquely the scattering type with three parameters; the symmetric scattering type magnitude alphas and phase phis introduced in  and the target helicity [5, 6]. In this study, the Touzi decomposition is investigated for wetland characterization using polarimetric Convair-580 SAR data collected over the RAMSAR Mer Bleue wetland in the East of Ottawa (Canada). It is shown that, like the Cloude alpha, the symmetric scattering type magnitude alphas is not effective for high entropy vegetation type discrimination. The phase phis of the symmetric scattering type has to be used for enhanced discrimination of wetland vegetation species and forest types. The use of the dominant scattering type phase, phis1, makes possible the discrimination of shrub-bog from sedges-fen. These two classes cannot be separated using like channel phase difference PhiHH-PhiVV or the radiometric scattering information provided by alphas, the Cloude alpha, the entropy H, or the multi-polarization HH-HV-VV channels. Phis1 also permits discriminating conifer-dominated treed bog from upland deciduous forest under leafy conditions, and looks promising for the detection of spring versus fall wetland class changes. The analysis of low entropy marsh scattering showed that both the symmetric scattering type magnitude and phase, alphas1 and phis1 , as well as the maximum polarization intensity of the dominant scattering m1 are needed for a better understanding of marsh complex scattering mechanisms. The use of medium scattering helicity is shown promising for characterization of forest structure.
 Touzi, R., Boerner, W., Lee, J.S., and Luneberg, E. A review of polarimetry in the context of synthetic aperture radar: Concepts and information extraction. Canadian Journal of Remote Sensing, Special issue on RADARSAT-2, pp. 380-407, June 2004.
 R. Cloude and E. Pottier, “A review of target decomposition theorems in radar polarimetry”, IEEE Trans. Geoscience Rem. Sens., Vol. 34, No. 2, pp. 498-518, March 1996.
 Touzi, R., “Target Scattering Decomposition in Terms of Roll-Invariant Target Parameters ”, Proc. IEEE TGRS, Vol. 45, No. 1, Jan. 2007.
 Touzi, R., “Speckle effect on polarimetric target scattering decomposition of SAR imagery”, Can. J. of Rem. Sens., Vol. 33, No. 1, pp. 60-68, Feb. 2007
 K. Kennaugh. Effects of Type of Polarization on Echo Characteristics. Technical report, The Ohio State University, Antenna Laboratory, Columbus, OH Report 389-4, 35p and 381-9, 39p , 1951.
 J.R. Huynen. Measurement of the target scattering matrix. Proc. IEEE,