CLASSIFICATION OF MULTIFREQUENCY POLARIMETRIC SAR IMAGERY USING A DYNAMIC LEARNING NEURAL-NETWORK

A practical method for extracting microwave backscatter for terrain-co ver classification is presented in this paper. The test data are multi frequency (P, L, C bands) polarimetric SBR data acquired by JPL over a n agricultural area called ''Flevoland.'' The terrain covers include f orest, water, bare soil, grass, and eight other types of crops. The ra dar response of crop types to frequency and polarization states were a nalyzed for classification based on three configurations: 1) multifreq uency and single-polarization images; 2) single-frequency and multipol arization images; and 3) multifrequency and multipolarization images. A recently developed dynamic learning neural network was adopted as th e classifier. Results show that using partial information, P-band mult ipolarization images and multiband hh polarization images, have better classification accuracy, while with a full configuration, namely, mul tiband and multipolarization, gives the best discrimination capability . The overall accuracy using the proposed method can be as high as 95% with a total of thirteen cover types classified. Further reduction of the data volume by means of correlation analysis was conducted to sin gle out the minimum data channels required. It was found that this met hod efficiently reduces the data volume while retaining highly accepta ble classification accuracy.