Novel Clustering Technique for Monitoring Crop Phenology Using SAR Data

Unsupervised clustering is essential when in-hand information about crop phenological stages are not available. In addition, this technique assists in discriminating crop growth stages concerning the changes in the scattering mechanisms. Newly proposed parameter $\theta_{FP}$ for full polarimetric Synthetic Aperture Radar (SAR) data is shown to be enhancing the target characterization capability over the existing parameter such as $\overline{\alpha}$. Following this, the existing clustering technique by combinedly utilizing the Barakat degree of polarization and elements of the Kennaugh matrix along with scattering entropy H
FP
is effective over land cover classes. The clustering scheme is consisting of 12 zones where zones (Zl, Z2, Z3), (Z10, Zll, Z12), and (Z4, Z5, Z6, Z7, Z8, Z9) represent even, odd and multiple scattering types, respectively. In this study, we use the RADARSAT-2 FP SAR data from Jun to Nov 2019 acquired over the Indian test site of Vijayawada under the Joint Experiment for Crop Assessment and Monitoring (JECAM) initiative. The changes and shifts of the clusters are exciting according to the phenological development of rice fields. At the initial stage, dense clusters are evident in Z10, Zll, and Z12, while, during the advanced tillering stage, more data points are found in the Z3 zone. Following this, clusters in zone Z5, Z6, Z8 and Z9 are evident during the maturity stage.