GEOSCAN Search Results: Fastlink

GEOSCAN Menu


TitleOptimization of the application of the Touzi decomposition for wetland classification using polarimetric Radarsat-2
AuthorGosselin, G; Touzi, R; Bhattacharya, A
Source33rd Canadian Symposium on Remote Sensing, abstracts; by Canadian Symposium on Remote Sensing; 2012 p. 12
LinksOnline - En ligne
LinksAbstracts (PDF, 1.22 MB)
Year2012
Alt SeriesEarth Sciences Sector, Contribution Series 20140075
Meeting33rd Canadian Symposium on Remote Sensing; Ottawa; CA; June 11-14, 2012
Documentbook
Lang.English
Mediaon-line; digital
File formatpdf
ProvinceQuebec
NTS31I/02; 31I/07
AreaLac Saint-Pierre; St. Lawrence River
Lat/Long WENS -73.0000 -72.5000 46.5000 46.0000
Subjectsgeophysics; remote sensing; satellite imagery; analytical methods; wetlands; Touzi decomposition; target scattering
AbstractTarget scattering decomposition has become the standard method for the extraction of natural target geophysical parameters from polarimetric SAR data (Cloude and Pottier, 1996; Touzi et al., 2004). In contrast to the Cloude-Pottier decomposition that characterizes target scattering type with a real entity, the so-called Cloude a, the Touzi decomposition (Touzi, 2007) uses the magnitude as and the phase Fas of the ''complex'' scattering type for unambiguous characterization of target scattering. Target helicity is also used to assess the symmetric nature of target scattering. To avoid any loss of information, no averaging of the scattering parameters is performed and target scattering parameters are characterized by an in-depth analysis of each of the three eigenvector parameters (12 parameters). All of the twelve scattering parameters might not be required for target scattering classification, and there is a need for the development of a procedure that permits the selection of the optimum parameter subset for an effective classification. Bhattacharya and Touzi (2012) have recently introduced a method based on the mutual information theory (the maximum relevancy and minimum redundancy (MRMR)) and the Support Vector Machines (SVM) that permits the selection of the optimum Touzi decomposition parameter subset. In this study, the MRMR-SVM method (Bhattacharya and Touzi 2012) is reconsidered. The effect of the parameter ranking on the classification is studied, and the loss of information related to the eventual limitation of the SVM in combining phase and intensity parameters is assessed. The investigation is conducted using a series of Radarsat-2 images collected in 2009 over the Lac Saint-Pierre region that includes urban, forest and wetland areas. The decomposition scattering multitemporal change is exploited to optimize the SVM-based classification.
GEOSCAN ID294581