Extended morphological profiles using auto-associative neural networks for hyperspectral data classification

Recently, morphological profiles have be observed as good tools to fuse spectral and spatial information to produce better classification results. In general, the profiles are built with the features derived using the principal component analysis (PCA). Auto-associative neural network (AANN), which can be seen as an implementation of non-linear PCA is used for unsupervised feature reduction of hyperspectral data. In this paper, we investigate the suitability of the features derived using AANN to build extended morphological profiles for hyperspectral data classification.