A concurrent neural network model for pattern recognition in multispectral satellite imagery

We investigate multispectral satellite image classification using the neural model previously proposed by the first author called Concurrent Self-Organizing Maps (CSOM), representing a winner-takes-all collection of self-organizing neural network modules. For comparison, we evaluate the performances of several statistical classifiers (Bayes, 1-NN, and K-means). The implemented neural versus statistical classifiers are evaluated using a LANDSAT ETM+ image composed by a set of 7-dimensional multispectral pixels, out of which a subset contains labeled pixels, corresponding to eleven thematic categories. The best experimental result leads to the recognition rate of 99.23 %.