Retrieving rice yield and biomass from Radarsat-2 SAR data with artificial neural network (ANN)

摘要

The main objective of this study was to retrieve rice yield and biomass from Radarsat-2 SAR data with artificial neural network (ANN). For this purpose, a practical scheme for estimating rice yield from Radarsat-2 data is established, which demonstrates that Radarsat-2 data can serve as an important data source for monitoring rice system and estimating rice yield. The ANN was composed of the rice backscattering coefficients extracted from multi-temporal Radarsat-2 images and rice canopy parameters (i.e. height, moisture content and biomass) observed from the fields, and then it was applied to simulate the correlation between these two parts. The rice yield and biomass on August 21 and September 14 were retrieved based on the trained network, respectively. Compared with the measured data, the retrieved rice yield and biomass on Aug. 21 and Sept. 14 were quite accurate. Our results suggested that Radarsat-2 SGX images can be used to estimate rice yield regionally, and neural network method is feasible with respects to the estimation of rice yield and biomass.