Forest characterization and mapping using fully polarimetric SAR data
Mounira Ouarzeddine(1), Aichouche Belhadj-Aissa(1) and Boularbah Souissi(1)
(1) USTHB, BP N°32 El Alia 16111, Bab Ezzouar, Alger, 16111, Algeria
The ability of radar multi-polarization to retrieve more information about physical properties of the surface has led to a wide range in geo-environmental applications. A particular attention has been given to forest applications such as clear cuts and linear features mapping, biomass estimation, species identification and fire scar mapping.
In this paper, the potential of the fully polarimetric SAR data in the improvement and the separability to characterize and discriminate different forest classes is investigated.
Firstly we have applied the modified four component model decomposition which is the extended Freeman three-component polarimetric decomposition to segment the image in different scattering types. Secondly The randomness of the medium represented by the entropy defined by cloude and Pottier has been introduced with the four scattering components to extend the classes from 4 to 12 depending on the degree of the randomness. Results of the extended decomposition were used as training set for the initialization of the Wishart classifier. The final combined classification gives more details on the different tree classes in the forested area. The polarimetric analysis and classification algorithms have been tested using the fully polarimetric data sets acquired in the region of Mer bleue in Canda using C band.