COMPARISON OF PROJECTION PURSUIT AND MATCHING PURSUIT IN FEATURE EXTRACTION OF HYPERSPECTRAL IMAGES

摘要

The purpose of feature extraction is to reduce the dimensionality of hyperspectral images that the problem of "curse of dimensionality" can be avoided. Two similar feature extraction methods based on the pursuit strategy are compared in this paper, they are projection pursuit and matching pursuit. The basic idea of projection pursuit is to locate the projection or projections from high- to low-dimensional space that reveal the most details about the structure of the data set. Matching pursuit is a general procedure for computing adaptive signal representation using only a few atoms from a wavelet dictionary. For the application of hyperspectral data, the wavelet packet dictionary with different deep of levels can be used to describe the local spectral information in different size of spectral wavelength intervals. Thus the matching pursuit can extract important features for hyperspectral image classification. In this paper, these two pursuit methods are firstly described. Then a hyperspectral data set is used to test the performance of these two methods. The maximum likelihood classifier is used to test the classification performance after the feature extraction. The experiment results showed that both of these two methods are effective for feature extraction. Especially the matching pursuit method can provide an effective way for the hyperspectral images classification with small number of features.