Loading [MathJax]/extensions/MathMenu.js
Feature Extraction for Hyperspectral Images Using Low-Rank Representation With Neighborhood Preserving Regularization | IEEE Journals & Magazine | IEEE Xplore

Feature Extraction for Hyperspectral Images Using Low-Rank Representation With Neighborhood Preserving Regularization


Abstract:

Hyperspectral images (HSIs) usually contain hundreds of spectral bands. When they are used for classification tasks, HSIs may suffer from the curse of high dimensionality...Show More

Abstract:

Hyperspectral images (HSIs) usually contain hundreds of spectral bands. When they are used for classification tasks, HSIs may suffer from the curse of high dimensionality. To address this problem, the essential procedures of dimension reduction and feature extraction (FE) are employed. In this letter, we propose an FE method for HSIs using low-rank representation with neighborhood preserving regularization (LRR_NP). The proposed method can simultaneously employ locally spatial similarity and the spectral space structure, which comprises a union of multiple low-rank subspaces. The framework of LRR can structurally represent the union structure of a spectral space. Because spatial neighbor pixels always share high similarity in a feature space, an NP regularization item is introduced into the framework of LRR to consider the locally spatial correlation. Classification experiments are conducted on real HSI data sets; the results demonstrate that the features that are extracted by LRR_NP are more discriminative than the state-of-art methods, including both unsupervised methods and supervised methods.
Published in: IEEE Geoscience and Remote Sensing Letters ( Volume: 14, Issue: 6, June 2017)
Page(s): 836 - 840
Date of Publication: 04 April 2017

ISSN Information:

Funding Agency:


I. Introduction

Because hyperspectral images (HSIs) provide not only spatial information but also spectral information about a scene, they are extensively employed in mineral identification, agriculture yield estimation, and military target detection. Due to the immense number of feature variables (spectral bands), the application of HSIs generally requires a large amount of memory and computation power [1]. As a result, reducing the dimensionality while simultaneously retaining the structural discriminative features is important for HSIs. Dimension reduction can be realized using two ways: feature selection (FS) and feature extraction (FE). FS tends to select the most discriminative features from the existing feature sets, whereas FE aims to find a transform to build the derived informative features. FS can only select from a previously existing feature set while discarding other feature sets, whereas FE can use all features and generate more discriminative features via certain computations. In this letter, we choose to employ FE to reduce the feature dimension in HSIs.

References

References is not available for this document.