Neural versus statistical approaches for pattern recognition in space imagery [articol]

Abstract

We investigate multispectral space image classification using the new neural model called Concurrent Self-Organizing Maps (CSOM), representing a winner-takes-all collection of small modular self-organizing neural networks. For comparison, we evaluate the performances of Bayes classifier. The implemented neural/statistical classifiers are evaluted using a LANDSAT TM image with 7 bands composed by a set of 7-dimensional pixels, out of which a subset contains labeled pixels, corresponding to seven thematic categories . The best experimental result leads to the recognition rate of 95.29 %.