Locality sensitive discriminant analysis for classification of hyperspectral data

For hyperspectral data classification, feature reduction techniques have become an apparent need to extract information from original data. In this paper, we introduce Locality Sensitive Discriminant Analysis (LSDA) to perform feature reduction for classification of hyperspectral imagery. By preserving both the discriminant and local geometrical structure in the data, the proposed method can obtain better classification accuracy with less computational load. Experiments carried out on two data sets, Indian Pine and Pavia City, show that results of our method outperform that of other traditional dimensionality reduction methods.