Polarimetric analysis of multi-temporal RADARSAT-2 SAR images for wheat monitoring and mapping

A full understanding of the polarimetric characteristics of wheat fields is necessary for the development of a robust methodology for the monitoring and mapping of wheat using quad-polarimetric SAR images. In this study, the polarimetric characteristics and temporal variations of wheat were analysed using a multi-temporal RADARSAT-2 quad-polarimetric dataset from a wheat-growing region of the North China Plain. The backscattering coefficient, the Freeman–Durden decomposition, and the H/A/[Inline formula] decomposition were evaluated as functions of the growth stage and then used for classification. With each wheat growth stage, the volume scattering component ratio increased, whereas the surface scattering component ratio generally decreased. The experimental results indicate that the Freeman–Durden decomposition parameters are sensitive to the wheat growth stage. Moreover, the proposed method for mapping wheat, which combines the backscattering coefficients, polarization decompositions, and the support vector machine (hereafter referred to as BP-SVM), is able to discriminate wheat effectively, with an accuracy of up to 92.92%. This indicates that quad-polarimetric imagery from just one date is sufficient for wheat mapping. The results of this study demonstrate that quad-polarimetric RADARSAT-2 SAR imagery has great potential for the monitoring and mapping of wheat.