A Gaussian adaptive resonance theory neural network classification algorithm applied to supervised land cover mapping using multitemporal vegetation index data

Neural network classifiers have been shown to provide supervised classifica tion results that significantly improve on traditional classification algor ithms such as the Bayesian (maximum likelihood [ML]) classifier. While the predominant neural network architecture has been the feedforward multilayer perceptron known as backpropagation, Adaptive resonance theory (ART) neura l networks offer advantages to the classification of optical remote sensing data for vegetation and land cover mapping. A significant advantage is tha t it does not require prior specification of the neural net structure, crea ting as many internal nodes as are needed to represent the calibration (tra ining) data. The Gaussian ARTMAP classification algorithm bases the probabi lity that input training samples belong to specific classes on the paramete rs of its Gaussian distributions: the means, standard deviations, and a pri ori probabilities. The performance of the Gaussian ARTMAP classification al gorithm in terms of classification accuracy using independent validation da ta indicated was over 70% accurate when applied to an annual series of mont hly 1-km advanced very high resolution radiometer (AVHRR) satellite normali zed difference vegetation index (NDVI) data. The accuracies were comparable to those of fuzzy ARTMAP and a univariate decision tree, and significantly higher than a Bayesian classification algorithm. Algorithm testing is base d on calibration and validation data developed and applied to Central Ameri ca to map the International Geosphere-Biosphere Programme (IGBP) land cover classification system. Thus, it provides a realistic test of the algorithm s for operational classification of a regional remote sensing and site data set.