Hyperspectral Image Classification Algorithm Based on Spectral Clustering and Sparse Representation

Anguo Dong; Jiaxun Li; Bei Zhang; Miaomiao Liang

doi:10.3788/AOS201737.0828005

Journals >Acta Optica Sinica >Volume 37 >Issue 8 >Page 0828005 > Article

Acta Optica Sinica
Vol. 37, Issue 8, 0828005 (2017)

Hyperspectral Image Classification Algorithm Based on Spectral Clustering and Sparse Representation

Anguo Dong¹, Jiaxun Li^1、*, Bei Zhang¹, and Miaomiao Liang²

Author Affiliations

¹ School of Science, Chang'an University, Xi'an, Shaanxi 710064, China

² School of Electronic Engineering, Xidian University, Xi'an, Shaanxi 710071, China

show less

DOI: 10.3788/AOS201737.0828005 Cite this Article Set citation alerts

Anguo Dong, Jiaxun Li, Bei Zhang, Miaomiao Liang. Hyperspectral Image Classification Algorithm Based on Spectral Clustering and Sparse Representation[J]. Acta Optica Sinica, 2017, 37(8): 0828005 Copy Citation Text

show less

Fig. 1. Hyperspectral images. (a) Category of real ground; (b) classification result of OMP algorithm; (c) spectral curves

Download full size

Fig. 2. Clustering results of ground in the neighborhood. (a) (129,35); (b) (96,39); (c) (38,52); (d) (100,58)

Download full size

Fig. 3. Contrast figures before and after algorithm correction. (a) Before correction; (b) after correction

Download full size

Fig. 4. Classification results of Pavia University dataset obtained by different algorithms. (a) Original image; (b) real ground; (c) SVM algorithm; (d) CK-SVM algorithm; (e) OMP algorithm; (f) SOMP algorithm; (g) MASR algorithm; (h) SC-SOMP algorithm

Download full size

Fig. 5. Classification results of Indian Pines dataset obtained by different algorithms. (a) Original image; (b) real ground; (c) SVM algorithm; (d) CK-SVM algorithm; (e) OMP algorithm; (f) SOMP algorithm; (g) MASR algorithm; (h) SC-SOMP algorithm

Download full size

Fig. 6. Classification results of Salinas Valley dataset obtained by different algorithms. (a) Original image; (b) real ground; (c) SVM algorithm; (d) CK-SVM algorithm; (e) OMP algorithm; (f) SOMP algorithm; (g) MASR algorithm; (h) SC-SOMP algorithm

Download full size

Fig. 7. Effect of the number of training samples. (a) Pavia University; (b) Indian Pines; (c) Salinas Valley

Download full size

Class	Sample	Classification algorithm
Class	Train		Test	SVM			CK-SVM	OMP		SOMP	MASR	SC-SOMP
Asphalt	250	6381			80.50	97.90	49.72	66.27	77.26	91.87
Meadows	250	18339			84.48	98.95	62.36	92.32	96.62	99.11
Gravel	250	1849			78.91	93.77	63.00	96.76	99.18	99.78
Trees	250	2814			96.24	98.96	84.49	94.94	96.91	98.33
Painted metal sheets	250	1095			99.74	100.00	99.12	99.13	100.00	99.36
Bare soil	250	4779			83.96	97.06	54.12	92.73	98.74	100.00
Bitumen	250	1080			91.39	99.56	83.46	99.16	99.99	99.91
Self-blocking bricks	250	3432			81.27	96.44	62.27	90.06	96.18	98.08
Shadows	250	697			98.44	99.87	95.18	85.02	83.59	89.67
OA /%					84.98	98.16	63.55	88.77	93.86	98.03
Kappa					0.80	0.98	0.54	0.85	0.92	0.97

Table 1. Experimental data and classification accuracies of the Pavia University dataset

Class	Classification algorithm
Class	SVM	CK-SVM	OMP	SOMP	MASK	SC-SOMP
OA /%	77.64	94.86	61.01	90.88	98.41	98.37
Kappa	0.74	0.94	0.66	0.90	0.98	0.97

Table 2. Classification accuracies of Indian Pines dataset obtained by different algorithms

Class	Classification algorithm
Class	SVM	CK-SVM	OMP	SOMP	MASR	SC-SOMP
OA /%	86.29	94.56	82.17	88.81	88.04	98.34
Kappa	0.85	0.94	0.8	0.88	0.87	0.97

Table 3. Classification accuracies of Salinas Valley dataset obtained by different algorithms

Anguo Dong, Jiaxun Li, Bei Zhang, Miaomiao Liang. Hyperspectral Image Classification Algorithm Based on Spectral Clustering and Sparse Representation[J]. Acta Optica Sinica, 2017, 37(8): 0828005

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information