Optimal datasets suitability for pearl millet (Bajra) discrimination using multiparametric SAR data

Abstract

In view with the importance of coarse cereals, it was proposed to discriminate pearl millet crop in a mixed crop scenario using different combinations of time series multipolarized SAR with available optical data. Five date dual polarized Sentinel-1 DT classification was found to be optimum, as Bajra classification accuracy increased by 7.6% over four date. The improved Bajra accuracy of 85% was achieved when 2 date Sentinel-1was integrated with higher value quad polarized Radarsat-2 data. Similar structured crops like Sorghum and maize could be discriminated using difference in crop calendar and polarization response. Single date Sentinel-2 (near harvest stage as available) with 2 date Sentinel-1 yields 66% accuracy for Bajra crop but the attempt to extract temporal phenological profile was failed due to persistent cloud cover in the previous dates. Radarsat-2 data increased Bajra accuracy up to 89.5% when responsive polarimetric parameters were used in conjunction with multiple linear polarizations.