Recognition of urban structures in multiband data by means of ART networks

In this paper multiband images of a urban environment are analyzed and interpreted by means of a neural network approach. In particular, the advantages found by using Adaptive Resonance Theory networks both for a spatial and spectral analysis of the data are shown and commented. Moreover, we simplify existing similar approaches by introducing a clustering step that automatically solves the problem of class redundancy, typical of the ART classification output. Results are given for a photo + SAR image of Santa Monica, Los Angeles.