Abstract
Sparse representation can be used for the representation of high-dimensional data into a low-dimensional subspace. Recently, sparse graph-based discriminant analysis that uses l 1-norm optimization has drawn much attention in dimensionality reduction of hyperspectral imagery. By combining low-rankness and sparsity, sparse and low-rank representation based discriminant analysis (SLGDA) can effectively capture global and local data structures simultaneously. In this article, collaborative and low-rank representation based discriminant analysis (CLGDA) is proposed, which is different from the concept of sparse representation. The more informative graph can be obtained in CLGDA with the combination of both collaborative representation (CR) and low-rank representation (LRR) because global data structure can be preserved by LRR and collaboration among within-class atoms is important than competition in sparse representation. Moreover, CR with l 2-norm regularization is computationally efficient and competitive to sparse representation. The experimental results conducted on three hyperspectral datasets demonstrate that the proposed CLGDA and its variants can provide better classification performance in comparison to SLGDA counterparts with lower computational complexity.
Highlights
H YPERSPECTRAL imagery (HSI) acquired through airborne or spaceborne sensors consist of hundreds of spectral bands for each pixel, resulting in a 3-D data cube
In order to improve the performance of sparsitypreserving GE (SPGE) that can be considered as an unsupervised dimensionality reduction (DR) method, a supervised version of SPGE using class label information is introduced in [16] that is called as sparse graph-based discriminant analysis (SGDA) for DR
We propose collaborative and low-rank graphbased discriminant analysis (CLGDA) for DR of HSI
Summary
H YPERSPECTRAL imagery (HSI) acquired through airborne or spaceborne sensors consist of hundreds of spectral bands for each pixel, resulting in a 3-D data cube. Advanced projection-based techniques are developed, such as kernel LDA [10], kernel PCA [22], neighborhood preserving embedding [11], locally linear embedding [23], locality preserving embedding [12], local Fisher discriminant analysis (LFDA) [24], and kernel version of LFDA [25]. In order to improve the performance of SPGE that can be considered as an unsupervised DR method, a supervised version of SPGE using class label information is introduced in [16] that is called as sparse graph-based discriminant analysis (SGDA) for DR. We propose collaborative and low-rank graphbased discriminant analysis (CLGDA) for DR of HSI. SHAH AND DU: COLLABORATIVE AND LOW-RANK GRAPH FOR DISCRIMINANT ANALYSIS OF HYPERSPECTRAL IMAGERY in CLGDA can help to reduce computational cost.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
