Application of texture to JERS-1 SAR imagery for tropical forest land cover classification

Shows the potential of the Japanese Earth Remote Sensing Satellite (JERS-1) at L band and HH polarization 100-metre resolution SAR data towards identifying major land cover types of tropical rainforests. Texture statistics has been used as the feature dimensions in maximizing class discrimination. For this purpose, eight of the most common first order texture measurements found in the literature were used. Their functional analysis was based on histogram plots, scatter plots and separability tests using the Bhattachurrya distance. The results were then input into a maximum likelihood classifier and the classified image was validated as well as visually assessed. The study sites have been chosen over the Amazon Basin and the land cover classes are forest, nonforest, flooded vegetation and water.