Rice field mapping and monitoring using ASAR data based on principal component analysis

Synthetic Aperture Radar (SAR) is anticipated to be the dominant remote sensing data source for rice inventory in tropical and subtropical regions due to its independent from cloud cover. With multi-polarization SAR data, the accuracy for rice mapping may be increased. Multi-temporal ENVISAT ASAR alternative polarization data were used for the identification of rice crop in Fuzhou, Fujian Province. After analysis of the backscatter calculated from ASAR data, it showed that the rice backscatter increased with rice growing, and the backscatter of vertical polarization (VV) was larger than that of vertical and horizontal cross polarization (VH). In the late period of rice growing stage, the backscatter of VV kept stable, while the VH increased. Principal component transform was performed for three pairs of ASAR dual-polarization data. It was found that, in the 2nd component (PC2), the value of rice fields was high and showed very bright, while in the 5th component (PC5), the value of rice fields was low and showed in deep dark color, which mainly reflected the differences of rice field in early growing season and other growing seasons in VV and VH polarizations, respectively. The difference between PC2 and PC5 (PC2-PC5) improved the separability of rice and other land covers. Based on the difference of principal components (PC2-PC5), a method for rice field mapping was established using object oriented classifier. With this method, early rice fields of Fuzhou in 2004 were extracted much easily and quickly, and satisfying accuracy was obtained.