Eigen vector-based classification of pearl millet crop in presence of other similar structured (sorghum and maize) crops using fully polarimetric Radarsat-2 SAR data

Abstract

The effect of structure and phenology of dryland crops on SAR scattering mechanisms is under explored and under reported. This study aims to understand the complex scattering mechanisms occurring in similar structured crops and to discriminate pearl-millet crop using single date Radarsat-2 SAR data. The study arrived at few significant findings that the panicle stage isthe critical stage for discriminating pearl-millet from similar structured sorghum and maize crops. Alpha angle extracted from eigen-vector based decomposition was identified as the discriminating factor. Average alpha value almost reached 55° due to dihedral scattering occurring in the prominent panicles which emerge early in pearl millet compared to maize and sorghum. Wishart H-A-α technique discriminated pearl millet, maize and sorghum crops with a producer's accuracy of about 87.8%, 71.6% and 71.4%, respectively. Paddy was classified with highest accuracy of 95.8% due to occurrence of dominant double bounce scattering in most of the fields.