Fusion of Multiscale Low-rank Representation and Two Way Recursive Filtering for Hyperspectral Image Classification

Mei LU; Jiatian LI; Wen LI; Mihong HU; Jiaxin YANG

doi:10.11873/j.issn.1004-0323.2024.2.0393

Journals >Remote Sensing Technology and Application >Volume 39 >Issue 2 >Page 393 > Article

Remote Sensing Technology and Application
Vol. 39, Issue 2, 393 (2024)

Fusion of Multiscale Low-rank Representation and Two Way Recursive Filtering for Hyperspectral Image Classification

Mei LU^*, Jiatian LI, Wen LI, Mihong HU, and Jiaxin YANG

Author Affiliations

Faculty of Land Resource Engineering Kunming University of Science and Technology，Kunming 650000，China

show less

DOI: 10.11873/j.issn.1004-0323.2024.2.0393 Cite this Article

Mei LU, Jiatian LI, Wen LI, Mihong HU, Jiaxin YANG. Fusion of Multiscale Low-rank Representation and Two Way Recursive Filtering for Hyperspectral Image Classification[J]. Remote Sensing Technology and Application, 2024, 39(2): 393 Copy Citation Text

show less

Fig. 1. The flowchart to MSLRR_TWRF

Download full size | View in the Article

Fig. 2. Influence of parameter C on classification accuracy of MSLRR_TWRF method

Download full size | View in the Article

Fig. 3. Influence of parameter S_f on classification accuracy of MSLRR_TWRF method

Download full size | View in the Article

Fig. 4. Influence of parameter δ_r on classification accuracy of MSLRR_TWRF method

Download full size | View in the Article

Fig. 5. Indian Pines image classification results obtained by different methods

Download full size | View in the Article

Fig. 6. PaviaU image classification results obtained by different methods

Download full size | View in the Article

Fig. 7. Salinas image classification results obtained by different methods

Download full size | View in the Article

数据集

光谱范围

/nm

空间分辨

率/m

图像尺寸

波段数

类别数

带标签样本

Indian Pines

400~2 500

20×20

145×145

200

10 249

PaviaU

430~860

1.3×1.3

610×340

103

42 776

Salinas

400~2 500

3.7×3.7

512×217

204

54 129

Table 1. Dataset Description

View in the Article

Class	Train	Test	SVM	PCA	IFRF	HiFi	CCJSR	R-VCANet	SSRN	MSLRR	MSLRR_ TWRF
Alfalfa	10	36	38.81	23.61	77.45	98.61	67.24	100	90	98.33	100
Corn_n	10	1 418	50.69	52.56	75.68	67.48	63.88	49.53	74.76	71.09	70.92
Corn_m	10	820	42.22	39.41	62.20	84.87	64.30	73.49	73.47	75.51	79.52
Corn	10	227	27.06	25.06	54.89	92.73	41.34	94.63	86.64	96.21	96.78
Grass_m	10	473	75.37	60.97	88.19	78.48	95.78	88.92	95.86	85.67	86.96
Grass_t	10	720	84.91	79.21	91.15	96.60	95.82	94.47	93.33	99.64	98.03
Grass_P	10	18	39.69	26.64	50.84	98.89	37.46	100	79.82	97.78	97.78
Hay_w	10	468	96.20	97.33	100	96.88	99.57	95.73	91.75	100	100
Oats	10	10	14.18	21.27	30.77	100	13.77	100	100	100	100
Soybean_n	10	962	52.26	39.96	69.67	85.30	62.36	76.17	77.54	84.83	86.75
Soybean_m	10	2 445	66.89	63.46	87.61	70.49	78.69	65.38	83.93	89.91	91.90
Soybean_c	10	583	33.15	33.63	77.36	84.15	68.03	71.90	79.80	87.32	91.56
Wheat	10	195	79.56	78.75	76.86	99.33	90.40	99.18	98.08	99.49	99.49
Woods	10	1 255	91.32	87.97	97.97	93.64	96.56	91.68	93.45	90.91	97.10
Buildings	10	376	38.23	35.28	78.70	89.65	73.62	81.49	83.29	91.65	92.69
stone	10	83	83.48	84.72	96.75	99.28	90.75	99.52	97.02	95.78	97.95
OA	/	/	59.44±0.03	54.60±0.03	79.14±0.03	81.97±0.02	74.93±0.02	75.49±0.01	80.63±0.04	87.04±0.03	89.05±0.02
AA	/	/	57.13±0.03	53.11±0.02	76.01±0.05	89.77±0.01	71.22±0.2	86.38±0.01	87.42±0.02	91.51±0.09	92.96±0.08
Kappa	/	/	54.47±0.03	49.27±0.04	76.51±0.04	79.69±0.03	71.76±0.02	72.44±0.03	77.97±0.05	85.20±0.04	87.48±0.03

Table 2. Classification accuracy of different methods for Indian Pines

View in the Article

Class	Train	Test	SVM	PCA	IFRF	HiFi	CCJSR	R-VCANet	SSRN	MSLRR	MSLRR_ TWRF
Asphalt	10	18 639	88.79	66.01	69.94	72.08	90.98	70.93	93.22	84.87	88.77
Meadows	10	2 089	84.09	65.53	94.43	79.36	84.62	76.75	95.44	92.16	95.44
Gravel	10	3 054	46.35	57.32	61.05	78.44	42.07	87.18	80.10	94.98	98.23
Trees	10	1 335	54.04	91.72	53.87	78.79	67.26	93.74	98.29	93.15	88.87
Sheets	10	5 019	85.41	99.75	97.48	89.62	66.77	99.97	98.69	97.69	99.57
Soil	10	1 320	36.31	56.21	77.53	77.30	35.68	85.62	94.77	84.36	98.74
Bitumen	10	3 672	42.66	88.58	66.62	92.73	60.25	93.38	90.82	96.11	99.96
Bricks	10	937	70.29	72.46	58.13	72.57	45.61	81.27	79.07	92.08	97.27
Shadows	10	180	98.85	99.64	49.96	99.01	75.25	98.34	99.83	99.45	99.66
OA	/	/	64.87±0.05	69.11±0.05	74.53±0.06	78.48±0.05	61.05±0.04	80.72±0.023	92.35±0.01	90.77±0.02	94.98±0.02
AA	/	/	67.42±0.04	77.14±0.18	69.89±0.04	82.21±0.03	63.17±0.2	87.46±0.011	92.25±0.02	92.76±0.05	96.28±0.04
Kappa	/	/	56.54±0.06	61.40±0.05	67.85±0.06	72.50±0.05	50.82±0.04	75.52±0.03	89.90±0.02	87.91±0.02	93.42±0.03

Table 3. Classification accuracy of different methods for PaviaU

View in the Article

Class	Test	Train	SVM	PCA	IFRF	HiFi	CCJSR	R-VCANet	SSRN	MSLRR	MSLRR_ TWRF
Weeds_1	10	1 999	97.60	97.18	97.03	99.45	99.95	99.87	87.18	100	100
Weeds_2	10	3 716	98.93	98.18	99.98	99.46	99.47	99.64	99.82	99.68	100
fallow	10	1 966	88.34	90.35	99.81	98.11	93.51	97.22	95.06	89.35	100
fallow_P	10	1 384	98.31	98.57	91.47	98.92	99.17	99.55	98.73	98.66	98.36
Fallow_s	10	2 668	95.96	97.01	99.63	97.75	93.04	99.57	96.72	96.63	98.34
stubble	10	3 949	99.82	99.88	99.92	97.68	93.53	99.80	99.88	99.73	99.95
Celery	10	3 569	95.73	93.21	98.89	98.83	98.86	98.94	99.92	99.59	99.92
Grapes	10	11 261	71.62	68.56	97.48	66.37	76.58	70.89	82.71	80.99	98.43
Soil	10	6 193	99.47	97.74	99.99	99.76	98.52	98.86	99.32	98.02	100
Corn	10	3 268	79.98	86.71	99.66	84.39	95.65	86.16	96.19	88.28	98.27
Lettuce_4	10	1 058	83.00	87.09	98.14	94.04	91.00	95.43	95.12	96.77	100
Lettuce_5	10	1 917	91.84	86.98	97.65	99.93	95.62	99.95	97.22	95.33	98.59
Lettuce_6	10	906	88.75	88.49	92.71	99.44	77.55	98.79	97.31	98.13	97.99
Lettuce_7	10	1 060	90.98	95.75	91.65	95.14	94.15	95.25	98.67	96.07	98.51
Vinyard_U	10	7 258	49.80	49.12	79.70	76.84	53.42	77.62	75.55	62.25	99.37
Vinyard_T	10	1 797	97.20	92.22	99.96	92.75	99.57	91.93	99.88	97.39	99.91
OA	/	/	82.71±0.02	82.16±0.02	95.07±0.01	87.92±0.03	84.94±0.02	89.23±0.01	95.07±0.03	88.94±0.01	99.23±0.01
AA	/	/	89.21±0.01	89.19±0.01	96.48±0.01	93.68±0.01	91.22±0.1	94.34±0.01	94.96±0.03	93.55±0.1	99.23±0.08
Kappa	/	/	80.85±0.02	80.20±0.02	94.52±0.02	86.61±0.03	83.29±0.02	88.04±0.01	89.50±0.04	87.66±0.01	99.15±0.04

Table 4. Classification accuracy of different methods for Salinas

View in the Article

方法	指标	训练样本数
方法	指标	10	20	30	40	50
SVM	OA/%	59.44	68.60	73.25	75.97	78.56
	AA/%	57.13	64.52	70.33	73.25	75.85
	Kappa/%	54.47	64.56	69.71	72.75	75.60
PCA	OA/%	54.60	62.37	67.67	70.25	72.27
	AA/%	53.11	59.89	65.17	68.07	71.23
	Kappa/%	49.27	57.73	63.51	66.33	68.53
IFRF	OA/%	79.14	89.85	93.40	95.33	96.21
	AA/%	76.01	85.70	92.35	93.82	95.59
	Kappa/%	76.51	88.47	92.47	94.65	95.66
HiFi	OA/%	81.97	88.90	91.56	93.02	94.36
	AA/%	89.77	94.01	95.43	96.34	96.90
	Kappa/%	79.69	87.41	90.38	92.02	93.54
CCJSR	OA/%	74.93	81.57	86.03	88.17	89.45
	AA/%	71.22	75.53	78.73	79.34	80.67
	Kappa/%	71.76	79.09	84.15	86.23	87.95
R-VCANet	OA/%	75.49	83.23	86.72	89.16	91.05
	AA/%	86.38	91.33	94.01	95.08	95.92
	Kappa/%	72.44	80.99	84.95	87.67	90.12
SSRN	OA/%	80.63	89.87	94.92	95.34	97.37
	AA/%	87.42	84.55	90.18	85.19	91.57
	Kappav	77.97	88.45	94.19	94.70	97.41
MSLRR	OA/%	87.04	92.75	94.42	94.71	95.08
	AA/%	91.51	94.98	96.22	96.58	96.63
	Kappa/%	85.20	91.70	93.60	93.93	94.35
MSLRR_TWRF	OA/%	89.05	94.95	96.82	97.66	98.00
	AA/%	92.96	96.71	97.93	98.40	98.59
	Kappa/%	87.48	94.21	96.36	97.32	97.70

Table 5. Classification accuracy of different training samples for Indian Pines

View in the Article

方法	指标	训练样本数
方法	指标	10	20	30	40	50
SVM	OA/%	64.87	75.92	78.44	82.11	83.19
	AA/%	67.42	75.49	76.49	79.75	80.36
	Kappa/%	56.54	69.28	72.38	76.89	78.24
PCA	OA/%	69.11	76.38	77.48	79.19	81.29
	AA/%	77.47	81.82	83.54	84.86	86.32
	Kappa/%	61.40	69.84	71.38	73.50	76.04
IFRF	OA/%	74.53	86.48	89.87	92.83	93.92
	AA/%	69.89	81.89	85.90	89.05	90.74
	Kappa/%	67.85	82.50	86.75	90.57	91.97
HiFi	OA/%	78.48	85.41	87.52	88.87	89.92
	AA/%	82.21	88.29	90.29	91.75	91.90
	Kappa/%	72.50	81.18	83.81	85.55	86.68
CCJSR	OA/%	61.05	64.65	70.16	75.91	76.23
	AA/%	63.17	63.93	68.49	71.60	72.82
	Kappa/%	50.82	55.59	61.35	68.63	69.04
R-VCANet	OA/%	80.72	86.97	90.67	92.07	93.53
	AA/%	87.46	90.97	93.61	94.35	95.36
	Kappa/%	75.52	83.17	87.83	89.62	91.58
SSRN	OA/%	92.35	94.43	96.49	96.60	98.09
	AA/%	92.25	94.04	96.31	96.67	98.01
	Kappa/%	89.90	92.60	95.32	95.48	97.47
MSLRR	OA/%	90.77	94.68	95.50	96.24	96.78
	AA/%	92.76	95.44	96.22	96.91	97.34
	Kappa/%	87.91	92.98	94.06	95.03	95.75
MSLRR_TWRF	OA/%	94.98	97.43	97.94	98.14	98.31
	AA/%	96.28	97.99	98.40	98.67	98.84
	Kappa/%	93.42	96.60	97.28	97.54	97.76

Table 6. Classification accuracy of different training samples for PaviaU

View in the Article

方法	指标	训练样本数
方法	指标	10	20	30	40	50
SVM	OA/%	82.71	84.51	85.52	87.24	87.36
	AA/%	89.21	90.46	91.77	92.72	92.94
	Kappa/%	80.85	82.81	83.93	85.81	85.94
PCA	OA/%	82.16	84.98	85.71	86.31	87.05
	AA/%	89.19	90.99	91.98	92.41	92.94
	Kappa/%	80.20	83.32	84.13	84.79	85.60
IFRF	OA/%	95.07	97.98	98.30	98.64	99.05
	AA/%	96.48	98.46	98.86	99.01	99.20
	Kappa/%	94.52	97.76	98.11	98.49	98.95
HiFi	OA/%	87.90	91.33	91.79	92.71	93.09
	AA/%	93.68	95.64	96.07	96.59	96.86
	Kappa/%	86.61	90.37	90.88	91.88	92.31
CCJSR	OA/%	84.94	86.34	89.35	89.92	90.20
	AA/%	91.22	91.97	94.09	94.36	94.44
	Kappa/%	83.29	84.85	88.17	88.76	89.11
R-VCANet	OA/%	89.23	90.39	91.83	93.27	94.25
	AA/%	94.34	95.63	96.31	97.06	97.86
	Kappa/%	88.04	89.32	90.92	92.52	93.69
SSRN	OA/%	90.57	89.83	92.49	95.89	96.94
	AA/%	94.96	96.11	97.34	97.72	98.50
	Kappa/%	89.50	88.75	91.68	95.42	96.58
MSLRR	OA/%	88.94	90.73	90.92	91.74	92.00
	AA/%	93.55	95.03	95.35	95.91	96.10
	Kappa/%	87.68	89.65	89.88	90.79	91.08
MSLRR_TWRF	OA/%	99.23	99.42	99.68	99.74	99.77
	AA/%	99.23	99.44	99.58	99.67	99.70
	Kappa/%	99.15	99.35	99.64	99.71	99.74

Table 7. Classification accuracy of different training samples for Salinas

View in the Article

方法	SVM	PCA	IFRF	HiFi	CCJSR
Indian Pines	4.96	3.59	3.63	75.44	64.94
PaviaU	4.20	1.53	7.75	44.37	155.75
Salinas	11.43	7.03	6.60	48.08	357.59
方法	R-VCANet	SSRN	MSLRR	MSLRR_TWRF
Indian Pines	4 785.82	947.78	310.32	330.06
PaviaU	16 570.74	194.92	2 142.00	2 153.36
Salinas	18 093.95	959.88	1 153.38	1 170.84

Table 8. Running time of different compared methods

Download Citation

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享