Abstract

Band selection is a commonly used approach for dimensionality reduction in hyperspectral imagery. Affinity propagation (AP), a new clustering algorithm, is addressed in many fields, and it can be used for hyperspectral band selection. However, this algorithm cannot get a fixed number of exemplars during the message-passing procedure, which limits its uses to a great extent. This paper proposes an adaptive AP (AAP) algorithm for semi-supervised hyperspectral band selection and investigates the effectiveness of distance metrics for improving band selection. Specifically, the exemplar number determination algorithm and bisection method are addressed to improve AP procedure, and the relations between selected exemplar numbers and preferences are established. Experiments are conducted to evaluate the proposed AAP-based band selection algorithm, and the results demonstrate that the proposed method outperforms other popular methods, with lower computational cost and robust results.

© 2012 Optical Society of America

Full Article  |  PDF Article

References

  • View by:
  • |
  • |
  • |

  1. G. F. Hughes, “On the mean accuracy of statistical pattern recognizers,” IEEE Trans. Inf. Theory 14, 55–63 (1968).
    [CrossRef]
  2. C.-I. Chang, Hyperspectral Imaging: Signal Processing Algorithm Design and Analysis (Wiley, 2009).
  3. D. Landgrebe, “Hyperspectral image data analysis,” IEEE Signal Process. Mag. 19, 17–28 (2002).
    [CrossRef]
  4. C.-I Chang and Q. Du, “Estimation of the number of spectrally distinct signal sources in hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 42, 608–619 (2004).
    [CrossRef]
  5. J. M. Bioucas-Dias and J. M. P. Nascimento, “Hyperspectral subspace identification,” IEEE Trans. Geosci. Remote Sens. 46, 2435–2445 (2008).
    [CrossRef]
  6. Q. Du and H. Yang, “Similarity-based unsupervised band selection for hyperspectral image analysis,” IEEE Geosci. Remote Sens. Lett. 5, 564–568 (2008).
    [CrossRef]
  7. C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
    [CrossRef]
  8. C.-I. Chang and S. Wang, “Constrained band selection for hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 44, 1575–1585 (2006).
    [CrossRef]
  9. A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
    [CrossRef]
  10. N. Keshava, “Distance metrics and band selection in hyperspectral processing with applications to material identification and spectral libraries,” IEEE Trans. Geosci. Remote Sens. 42, 1552–1565 (2004).
    [CrossRef]
  11. A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
    [CrossRef]
  12. H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
    [CrossRef]
  13. C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
    [CrossRef]
  14. B. J. Frey and D. Dueck, “Clustering by passing messages between data points,” Science 315, 972–976 (2007).
    [CrossRef]
  15. B. J. Frey and D. Dueck, “Response to comment on ‘Clustering by passing messages between data points,’” Science 319, 726 (2008).
    [CrossRef]
  16. S. Jia, Y. T. Qian, and Z. Ji, “Band selection for hyperspectral imagery using affinity propagation,” in Digital Image Computing: Techniques and Applications Conference (IEEE, 2008), pp. 137–141.
  17. S. Jia, Z. Ji, and Y. T. Qian, “Band selection based hyperspectral unmixing,” in International Workshop on Imaging Systems and Techniques (IEEE, 2009), pp. 303–306.
  18. C. J. C. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining Knowledge Discovery 2, 121–167 (1998).
    [CrossRef]
  19. C.-I. Chang, Hyperspectral Imaging: Techniques for Spectral Detection and Classification (Kluwer, 2003).

2011 (2)

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
[CrossRef]

2009 (1)

A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
[CrossRef]

2008 (3)

J. M. Bioucas-Dias and J. M. P. Nascimento, “Hyperspectral subspace identification,” IEEE Trans. Geosci. Remote Sens. 46, 2435–2445 (2008).
[CrossRef]

Q. Du and H. Yang, “Similarity-based unsupervised band selection for hyperspectral image analysis,” IEEE Geosci. Remote Sens. Lett. 5, 564–568 (2008).
[CrossRef]

B. J. Frey and D. Dueck, “Response to comment on ‘Clustering by passing messages between data points,’” Science 319, 726 (2008).
[CrossRef]

2007 (2)

B. J. Frey and D. Dueck, “Clustering by passing messages between data points,” Science 315, 972–976 (2007).
[CrossRef]

A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
[CrossRef]

2006 (1)

C.-I. Chang and S. Wang, “Constrained band selection for hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 44, 1575–1585 (2006).
[CrossRef]

2004 (2)

C.-I Chang and Q. Du, “Estimation of the number of spectrally distinct signal sources in hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 42, 608–619 (2004).
[CrossRef]

N. Keshava, “Distance metrics and band selection in hyperspectral processing with applications to material identification and spectral libraries,” IEEE Trans. Geosci. Remote Sens. 42, 1552–1565 (2004).
[CrossRef]

2002 (1)

D. Landgrebe, “Hyperspectral image data analysis,” IEEE Signal Process. Mag. 19, 17–28 (2002).
[CrossRef]

1999 (1)

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

1998 (1)

C. J. C. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining Knowledge Discovery 2, 121–167 (1998).
[CrossRef]

1968 (1)

G. F. Hughes, “On the mean accuracy of statistical pattern recognizers,” IEEE Trans. Inf. Theory 14, 55–63 (1968).
[CrossRef]

Althouse, M. L. G.

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

Bioucas-Dias, J. M.

J. M. Bioucas-Dias and J. M. P. Nascimento, “Hyperspectral subspace identification,” IEEE Trans. Geosci. Remote Sens. 46, 2435–2445 (2008).
[CrossRef]

Burges, C. J. C.

C. J. C. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining Knowledge Discovery 2, 121–167 (1998).
[CrossRef]

Cariou, C.

C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
[CrossRef]

Chang, C.-I

C.-I Chang and Q. Du, “Estimation of the number of spectrally distinct signal sources in hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 42, 608–619 (2004).
[CrossRef]

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

Chang, C.-I.

C.-I. Chang and S. Wang, “Constrained band selection for hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 44, 1575–1585 (2006).
[CrossRef]

C.-I. Chang, Hyperspectral Imaging: Signal Processing Algorithm Design and Analysis (Wiley, 2009).

C.-I. Chang, Hyperspectral Imaging: Techniques for Spectral Detection and Classification (Kluwer, 2003).

Chehdi, K.

C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
[CrossRef]

Du, Q.

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

Q. Du and H. Yang, “Similarity-based unsupervised band selection for hyperspectral image analysis,” IEEE Geosci. Remote Sens. Lett. 5, 564–568 (2008).
[CrossRef]

C.-I Chang and Q. Du, “Estimation of the number of spectrally distinct signal sources in hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 42, 608–619 (2004).
[CrossRef]

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

Dueck, D.

B. J. Frey and D. Dueck, “Response to comment on ‘Clustering by passing messages between data points,’” Science 319, 726 (2008).
[CrossRef]

B. J. Frey and D. Dueck, “Clustering by passing messages between data points,” Science 315, 972–976 (2007).
[CrossRef]

Frey, B. J.

B. J. Frey and D. Dueck, “Response to comment on ‘Clustering by passing messages between data points,’” Science 319, 726 (2008).
[CrossRef]

B. J. Frey and D. Dueck, “Clustering by passing messages between data points,” Science 315, 972–976 (2007).
[CrossRef]

Hughes, G. F.

G. F. Hughes, “On the mean accuracy of statistical pattern recognizers,” IEEE Trans. Inf. Theory 14, 55–63 (1968).
[CrossRef]

Ji, Z.

S. Jia, Y. T. Qian, and Z. Ji, “Band selection for hyperspectral imagery using affinity propagation,” in Digital Image Computing: Techniques and Applications Conference (IEEE, 2008), pp. 137–141.

S. Jia, Z. Ji, and Y. T. Qian, “Band selection based hyperspectral unmixing,” in International Workshop on Imaging Systems and Techniques (IEEE, 2009), pp. 303–306.

Jia, S.

S. Jia, Z. Ji, and Y. T. Qian, “Band selection based hyperspectral unmixing,” in International Workshop on Imaging Systems and Techniques (IEEE, 2009), pp. 303–306.

S. Jia, Y. T. Qian, and Z. Ji, “Band selection for hyperspectral imagery using affinity propagation,” in Digital Image Computing: Techniques and Applications Conference (IEEE, 2008), pp. 137–141.

Keshava, N.

N. Keshava, “Distance metrics and band selection in hyperspectral processing with applications to material identification and spectral libraries,” IEEE Trans. Geosci. Remote Sens. 42, 1552–1565 (2004).
[CrossRef]

Landgrebe, D.

D. Landgrebe, “Hyperspectral image data analysis,” IEEE Signal Process. Mag. 19, 17–28 (2002).
[CrossRef]

Le Moan, S.

C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
[CrossRef]

Martínez-Usó, A.

A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
[CrossRef]

Melgani, F.

A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
[CrossRef]

Nascimento, J. M. P.

J. M. Bioucas-Dias and J. M. P. Nascimento, “Hyperspectral subspace identification,” IEEE Trans. Geosci. Remote Sens. 46, 2435–2445 (2008).
[CrossRef]

Paoli, A.

A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
[CrossRef]

Pasolli, E.

A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
[CrossRef]

Pla, F.

A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
[CrossRef]

Qian, Y. T.

S. Jia, Y. T. Qian, and Z. Ji, “Band selection for hyperspectral imagery using affinity propagation,” in Digital Image Computing: Techniques and Applications Conference (IEEE, 2008), pp. 137–141.

S. Jia, Z. Ji, and Y. T. Qian, “Band selection based hyperspectral unmixing,” in International Workshop on Imaging Systems and Techniques (IEEE, 2009), pp. 303–306.

Sheng, Y.

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

Sotoca, J. M.

A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
[CrossRef]

Su, H.

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

Sun, T.

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

Wang, S.

C.-I. Chang and S. Wang, “Constrained band selection for hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 44, 1575–1585 (2006).
[CrossRef]

Yang, H.

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

Q. Du and H. Yang, “Similarity-based unsupervised band selection for hyperspectral image analysis,” IEEE Geosci. Remote Sens. Lett. 5, 564–568 (2008).
[CrossRef]

Data Mining Knowledge Discovery (1)

C. J. C. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining Knowledge Discovery 2, 121–167 (1998).
[CrossRef]

IEEE Geosci. Remote Sens. Lett. (3)

H. Su, H. Yang, Q. Du, and Y. Sheng, “Semi-supervised band clustering for dimensionality reduction of hyperspectral imagery,” IEEE Geosci. Remote Sens. Lett. 8, 1135–1139 (2011).
[CrossRef]

C. Cariou, K. Chehdi, and S. Le Moan, “BandClust: an unsupervised band reduction method for hyperspectral remote sensing,” IEEE Geosci. Remote Sens. Lett. 8, 565–569 (2011).
[CrossRef]

Q. Du and H. Yang, “Similarity-based unsupervised band selection for hyperspectral image analysis,” IEEE Geosci. Remote Sens. Lett. 5, 564–568 (2008).
[CrossRef]

IEEE Signal Process. Mag. (1)

D. Landgrebe, “Hyperspectral image data analysis,” IEEE Signal Process. Mag. 19, 17–28 (2002).
[CrossRef]

IEEE Trans. Geosci. Remote Sens. (7)

C.-I Chang and Q. Du, “Estimation of the number of spectrally distinct signal sources in hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 42, 608–619 (2004).
[CrossRef]

J. M. Bioucas-Dias and J. M. P. Nascimento, “Hyperspectral subspace identification,” IEEE Trans. Geosci. Remote Sens. 46, 2435–2445 (2008).
[CrossRef]

C.-I Chang, Q. Du, T. Sun, and M. L. G. Althouse, “A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens. 37, 2631–2641 (1999).
[CrossRef]

C.-I. Chang and S. Wang, “Constrained band selection for hyperspectral imagery,” IEEE Trans. Geosci. Remote Sens. 44, 1575–1585 (2006).
[CrossRef]

A. Paoli, F. Melgani, and E. Pasolli, “Clustering of hyperspectral images based on multiobjective particle swarm optimization,” IEEE Trans. Geosci. Remote Sens. 47, 4175–4188 (2009).
[CrossRef]

N. Keshava, “Distance metrics and band selection in hyperspectral processing with applications to material identification and spectral libraries,” IEEE Trans. Geosci. Remote Sens. 42, 1552–1565 (2004).
[CrossRef]

A. Martínez-Usó, F. Pla, and J. M. Sotoca et al., “Clustering-based hyperspectral band selection using information measures,” IEEE Trans. Geosci. Remote Sens. 45, 4158–4171 (2007).
[CrossRef]

IEEE Trans. Inf. Theory (1)

G. F. Hughes, “On the mean accuracy of statistical pattern recognizers,” IEEE Trans. Inf. Theory 14, 55–63 (1968).
[CrossRef]

Science (2)

B. J. Frey and D. Dueck, “Clustering by passing messages between data points,” Science 315, 972–976 (2007).
[CrossRef]

B. J. Frey and D. Dueck, “Response to comment on ‘Clustering by passing messages between data points,’” Science 319, 726 (2008).
[CrossRef]

Other (4)

S. Jia, Y. T. Qian, and Z. Ji, “Band selection for hyperspectral imagery using affinity propagation,” in Digital Image Computing: Techniques and Applications Conference (IEEE, 2008), pp. 137–141.

S. Jia, Z. Ji, and Y. T. Qian, “Band selection based hyperspectral unmixing,” in International Workshop on Imaging Systems and Techniques (IEEE, 2009), pp. 303–306.

C.-I. Chang, Hyperspectral Imaging: Techniques for Spectral Detection and Classification (Kluwer, 2003).

C.-I. Chang, Hyperspectral Imaging: Signal Processing Algorithm Design and Analysis (Wiley, 2009).

Cited By

OSA participates in CrossRef's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (10)

Fig. 1.
Fig. 1.

Washington, DC, Mall data.

Fig. 2.
Fig. 2.

Classification comparison with different band selection methods for DC Mall data.

Fig. 3.
Fig. 3.

Classification maps of DC Mall data (with twelve bands). (a) All bands (OA=0.9340), (b) AP (OA=0.8331), and (c) AAP (OA=0.9426)

Fig. 4.
Fig. 4.

Purdue campus data.

Fig. 5.
Fig. 5.

Classification comparison with different band selection methods for Purdue campus data.

Fig. 6.
Fig. 6.

Classification comparison with different metrics in the AAP for Purdue campus data.

Fig. 7.
Fig. 7.

Classification maps of Purdue campus data (with twelve bands). (a) All bands (OA=0.8870), (b) AP (OA=0.8821), and (c) AAP (OA=0.9379).

Fig. 8.
Fig. 8.

Indian Pines data.

Fig. 9.
Fig. 9.

Amount of information with different band selection for Pines data.

Fig. 10.
Fig. 10.

Classification comparison with different band selection methods for the Indian Pines data.

Tables (3)

Tables Icon

Table 1. Computational Complexity of Band Selection Methodsa

Tables Icon

Table 2. Computing Time of Different Algorithms for Indian Pines Data (in Seconds)

Tables Icon

Table 3. Average Classification Accuracies Over 5–25 Selected Bands with Different Techniques

Equations (9)

Equations on this page are rendered with MathJax. Learn more.

s(i,k)=xixk2.
E(c)=i=1Ns(i,ci).
r(i,k)s(i,k)maxks.t.kk{a(i,k)+s(i,k)},
a(i,k)min{0,r(k,k)+is.t.i{i,k}max{0,r(i,k)}}.
a(k,k)is.t.ikmax{0,r(r,k)}.
c(i)=maxk(a(i,k)+r(i,k)).
M=λMold+(1λ)Mnew,
pref=h10i(hl).
pref=0.5h0.5l.

Metrics