new approach Automatic neuro - hyperspectral unmixing : a

and efficient solution to the unmixing problem. independence great discrimination ability on unlike signatures, giving a robust The model has great noise robustness, a correlation rate and endmember number always low Error Ratios for all cases. components, endmember number and proportion on the mixture, providing network behaviour vs. the Signal-to-Noise Ratio, correlation rate between order to use them for the rest of the experiments that consist on the study of the initially adjusting network parameters (learning parameter and stop criteria) in validating the model, studying the ratio error-value in the unmixing task, unmixing problem. Some types of experiments have been achieved for (abundances) from each endmember, giving one solution to the hyperspectral many output nodes as possible endmembers. It obtains the proportions model has one input node for each band of the hyperspectral signature and so based on the robustness and the inherent parallelism of NN. The proposed N N The use of non-supervised neural computation for hyperspectral analysis is basically consist of a mixture of radiance spectra. composition of earth canopies, where the pixels (hyperspectral signatures) automated fashion. Hyperspectral images are used to obtain the mineralogical the abundances of hyperspectral signatures obtained from remote sensors, in an The present work studies an Artificial Neural Network (NN) used to calculate