Hyperspectral Unmixing Method Based on Minimum Volume Sparse Regularization

Guangxian Xu; Yanwei Wang; Fei Ma; Feixia Yang

doi:10.3788/LOP57.241010

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 24 >Page 241010 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 24, 241010 (2020)

Hyperspectral Unmixing Method Based on Minimum Volume Sparse Regularization

Guangxian Xu¹, Yanwei Wang^1、*, Fei Ma^1、*, and Feixia Yang²

Author Affiliations

¹School of Electronic and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China

²School of Electrical and Control Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China

show less

DOI: 10.3788/LOP57.241010 Cite this Article Set citation alerts

Guangxian Xu, Yanwei Wang, Fei Ma, Feixia Yang. Hyperspectral Unmixing Method Based on Minimum Volume Sparse Regularization[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241010 Copy Citation Text

show less

Fig. 1. Linear mixed model of hyperspectral images

Download full size

Fig. 2. Minimum-volume simplex

Download full size

Fig. 3. Pseudo-code of MVSR-NMF algorithm

Download full size

Fig. 4. Pseudo-code of abundance features calculated by ADMM

Download full size

Fig. 5. Pseudo-code of endmember features calculated by ADMM

Download full size

Fig. 6. Elementary abundance maps of synthetic data SYN1.(a) Endmember 1; (b) endmember 2; (c) endmember 3; (d) endmember 4; (e) endmember 5; (f) endmember 6

Download full size

Fig. 7. Elementary abundance maps of synthetic data SYN2. (a) Endmember 1; (b) endmember 2; (c) endmember 3; (d) endmember 4; (e) endmember 5; (f) endmember 6

Download full size

Fig. 8. Urban scene image

Download full size

Fig. 9. Elementary abundance maps of synthetic data SYN1. (a) Original image; (b) FMVSA algorithm; (c) SISAL algorithm; (d) MVC-NMF algorithm; (e) CoNMF algorithm; (f) MVSR-NMF algorithm

Download full size

Fig. 10. Elementary abundance maps of synthetic data SYN2. (a) Original image; (b) FMVSA algorithm; (c) SISAL algorithm; (d) MVC-NMF algorithm; (e) CoNMF algorithm; (f) MVSR-NMF algorithm

Download full size

Fig. 11. Abundances of different algorithms. (a) Original image; (b) FMVSA algorithm; (c) SISAL algorithm; (d) MVC-NMF algorithm; (e) CoNMF algorithm; (f) MVSR-NMF algorithm

Download full size

Fig. 12. Comparison of endmember spectral bands under different algorithms. (a) Original endmember spectral bands; (b) endmember spectral bands of MVC-NMF algorithm; (c) endmember spectral bands of CoNMF algorithm; (d) endmember spectral bands of MVSR-NMF algorithm

Download full size

Algorithm	E_SA /(°)					E_FAA /(°)					t /s
Algorithm	20dB	25dB	30dB	35dB	40dB	20dB	25dB	30dB	35dB	40dB	t /s
FMVSA	6.69	4.63	2.61	1.63	1.03	22.93	16.28	9.34	5.52	3.19	21.55
SISAL	4.43	2.89	1.81	1.18	0.86	15.82	11.21	6.75	5.24	4.35	27.82
MVC-NMF	3.25	1.91	1.03	0.55	0.25	13.87	8.55	4.78	2.63	1.35	2159.97
CoNMF	1.93	1.46	1.02	0.51	0.28	12.47	7.97	5.22	3.28	1.96	127.22
MVSR-NMF	1.48	0.78	0.51	0.33	0.21	12.31	7.43	4.75	2.98	1.84	35.54

Table 1. Performance comparison of endmember estimation and abundance estimation of synthetic data SYN1in different signal-to-noise ratios

Algorithm	E_SA /(°)					E_FAA /(°)					t /s
Algorithm	20dB	25dB	30dB	35dB	40dB	20dB	25dB	30dB	35dB	40dB	t /s
FMVSA	6.82	6.33	2.60	1.61	0.99	30.25	29.35	16.26	10.64	6.37	23.48
SISAL	3.27	2.04	1.55	0.93	0.67	24.79	17.69	11.53	7.32	4.25	33.41
MVC-NMF	2.85	1.73	1.01	0.53	0.24	22.89	16.41	10.24	5.77	2.81	2549.52
CoNMF	1.68	1.13	0.89	0.49	0.35	18.97	14.45	9.98	5.94	3.68	244.27
MVSR-NMF	1.39	0.96	0.64	0.40	0.23	17.78	12.13	10.51	5.39	3.64	55.49

Table 2. Performance comparison of endmember estimation and abundance estimation of synthetic data SYN2 in different signal-to-noise ratios

Method	MVC-NMF	FMVSA	CoNMF	SISAL	MVSR-NMF
${\dot{E}}_{MS} (A, \dot{a})$	2.97	2.91	2.53	2.61	2.48
${\dot{E}}_{MS} (S, \dot{S})$	1.04	0.87	0.83	0.96	0.81

Table 3. Performance comparison of endmember matrix and abundance matrix in different algorithms

Guangxian Xu, Yanwei Wang, Fei Ma, Feixia Yang. Hyperspectral Unmixing Method Based on Minimum Volume Sparse Regularization[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241010

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information