GEOSCAN, résultats de la recherche


TitreUsing C-band fully polarimetric SAR data to discriminate orchard from forest stands
AuteurChalifoux, S; Charbonneau, F
SourceProceedings of the 14th Australasian Remote Sensing & Photogrammetry Conference, September 27 - October 4, 2008, Darwin, Australia; 2008, 9 pages
Séries alt.Secteur des sciences de la Terre, Contribution externe 20080214
RéunionProceedings of the 14th Australasian Remote Sensing & Photogrammetry Conference; Darwin; AU; Septembre 29 - Octobre 03, 2008
Medianumérique; CD-ROM
Sujetstélédétection; imagerie radar; méthodes radar; végétation; profils pédologiques; levés pédologiques; eau souterraine; cartographie par ordinateur; géomathématique
Illustrationsphotographs; 3-D images; digital images; Landsat images; satellite imagery; graphs
ProgrammeProgramme de cartographie des eaux souterraines
Résumé(disponible en anglais seulement)
The use of Earth observation data for land cover and land use mapping is a practical approach because of its extensive coverage and relatively low processing costs, comparative to traditional methods using aerial photographs or field surveys. Polarimetric synthetic aperture radar (SAR) and high resolution radar sensors offer researchers new possibilities to develop techniques for surface parameter mapping. Precise and current information about the distribution, composition and change in vegetation and soil surface is basic information for many terrestrial applications.
This research is part of an Earth observation application development activity, of the Groundwater Mapping Program at Natural Resources Canada, focusing on orchard (apple) detection and mapping to support the study of groundwater, specifically for aquifer recharge modeling and water demand analyses. Orchards, in a land cover mapping study using optical imagery, have a relatively low classification accuracy because of their confusion with agricultural lands and forest. In this activity, backscattering and decomposition algorithm analyses have been applied to a polarimetric dataset acquired by the C- band SAR onboard a Convair-580 aircraft owned by Environment Canada.
The experiment focuses on investigating the potentials of fully polarimetric SAR data to discriminate orchard from forest stands by analyzing backscattering signatures and decomposition techniques. Results show that orchards have lower backscattering intensity values at all polarizations compared to forested areas. Cloude and Pottier decomposition (entropy (H) and alpha angle (a)) was not successful to discriminate orchards from forested areas. Nevertheless, Freeman and Durden decomposition obtained interesting results. This technique permits the detection of orchard because of their relatively open areas and bare soils which is highlighted in backscattering by the ratio variation of double-bounce and diffusion scattering mechanisms. The supervised Wishart distribution was used to successfully classify orchard samples based on scattering categories.