Artificial neural network approach for vegetation classification from synthetic aperture radar images

This paper deals with the utilisation of artificial neural network to classify vegetation from highly nonlinear time varying backscatter parameters from the canopies and plants. The paper describes the backscatter phenomenon and their relevance with various types of plants and their constituents. An attempt is made to simulate and train an artificial NN package with the backscattering power experimentally obtained for two classes of vegetation, viz walnut orchard and coniferous forest, for a back propagation algorithm. The paper discusses the results achieved which is fairly accurate with reasonable elapsed time for the training. Further analysis of the simulated packages using Migraines software is underway