{5%, 10%, …, 30%}, the overall accuracy ranges from 99.95%
to 99.98%. Good classification accuracy can be obtained by
applying all proportions of training samples. When the patch
size is set as {1, 3, 5, 7, 9}, the corresponding overall accura-
cies are respectively {97.01%, 99.95%, 99.96%, 99.73%,
99.94%}. It can be seen that as long as the patch size is greater
than 3, the classification accuracy of the proposed
CSCSR
method is better than the accuracies of the other methods.
Conclusions
In this article, a
CSCSR
scheme is proposed for hyperspectral-
image classification. A testing pixel and its neighboring pixels
are projected onto a set of learned sparse representation vec-
tors. The class of the testing pixel is jointly determined by the
representation errors of pixels in the neighborhood. Experi-
mental results on three real-world
HSI
data sets demonstrate
the effectiveness of the proposed method. Since an alternative
method is used for searching
λ
1
and
λ
2
, the result may be a lo-
cal optimal solution. In future work, we will discuss the role
of these two parameters in more detail and seek the optimal
values of
λ
1
and
λ
2
. Also, we plan to investigate dimension-
ality reduction or segmentation techniques to reduce the
computational cost. We also plan to use the
CSCSR
method in
the field of
HSI
denoising.
Acknowledgments
This work was supported by the Provincial Natural Sci-
ence Foundation of Liaoning Province China, with Grants
2015020101 and 201602557, and the Scientific Research
Youth Project of Education Department of Liaoning Prov-
ince, China, Grant L201745. The authors also gratefully
acknowledge the helpful comments and suggestions of the
reviewers, which have improved the presentation.
References
Bioucas-Dias, J. M., A. Plaza, G. Camps-Valls, P. Scheunders, N.
Nasrabadi and J. Chanussot. 2013. Hyperspectral remote sensing
data analysis and future challenges.
IEEE Transactions on
Geoscience and Remote Sensing Magazine
1 (2):6–36.
Bioucas-Dias, J. M., A. Plaza, N. Dobigeon, M. Parente, Q. Du, P.
Gader and J. Chanussot. 2012. Hyperspectral unmixing overview:
Geometrical, statistical, and sparse regression-based approaches.
IEEE Journal of Selected Topics in Applied Earth Observations
and Remote Sensing
5 (2):354–379.
Camps-Valls, G. and L. Bruzzone. 2005. Kernel-based methods
for hyperspectral image classification.
IEEE Transactions on
Geoscience and Remote Sensing
43 (6):1351–1362.
Chen, C., W. Li, E. W. Tramel, M. Cui, S. Prasad and J. E. Fowler.
2014. Spectral-spatial preprocessing using multihypothesis
prediction for noise-robust hyperspectral image classification.
IEEE Journal of Selected Topics in Applied Earth Observations
and Remote Sensing
7 (4):1047–1059.
Chen, C., W. Li, E. W. Tramel and J. E. Fowler. 2014. Reconstruction
of hyperspectral imagery from random projections using
multihypothesis prediction.
IEEE Transactions on Geoscience
and Remote Sensing
52 (1):365–374.
Chen, Y., N. M. Nasrabadi and T. D. Tran. 2011. Hyperspectral image
classification using dictionary-based sparse representation.
IEEE
Transactions on Geoscience and Remote Sensing
49 (10):3973–
3985.
Chen, Y., N. M. Nasrabadi and T. D. Tran. 2013. Hyperspectral image
classification via kernel sparse representation.
IEEE Transactions
on Geoscience and Remote Sensing
51 (1):217–231.
Figure 18. Effect of the parameters on classification performance for the Salinas image: (a) effect of
λ
1
, (b) effect of
λ
2
, (c) effect
of the number of training samples, and (d) effect of the patch size.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2019
671
