Evaluation of Optical and Radar Images Integration Methods for LULC Classification in Amazon Region

The main objective of this study is to evaluate different methods to integrate (fusion and combination) Synthetic Aperture Radar (SAR) Advanced Land Observing Satellite (ALOS) Phased Arrayed L-band SAR (PALSAR-1) (Fine Beam Dual mode-FDB) and LANDSAT images in order to identify those which lead to higher accuracy of land-use and land-cover (LULC) mapping in an agricultural frontier region in Amazon. One method used to integrate the multipolarized information in SAR images before the fusion process was also evaluated. In this method, the first principal component (PC1) of SAR data was used. Color compositions of fused data that presented better LULC classification were visually analyzed. Considering the proposed objective, the following fusion methods must be highlighted: Ehlers, Wavelet á trous, Intensity, Hue and Saturation (IHS), and selective principal component analysis (SPC). These latter three methods presented good results when processed using PC1 from ALOS/PALSAR-1 FBD backscatter filtered image or three SAR extracted and selected features. These results corroborate with the applicability of the proposed method for SAR data information integration. Distinct methods better discriminate different LULC classes. In general, densely forested classes were better characterized by the Ehlers_TM6 fusion method, in which at least the polarization HV was used. Intermediate and initial regeneration classes were better discriminated using SPC-fused data with PC1 of ALOS/PALSAR-1 FBD data. Bare soil and pasture classes were better discriminated in optical features and the PC1 of ALOS/PALSAR-1 FBD data fused by the IHS method. Soybean with approximately 40 days from seeding was better discriminated in image classification obtained from ALOS/PALSAR-1 FBD image.