Hyperspectral Image Classification Based on Dual-Channel Feature Enhancement

Li Zhao; Leiquan Wang; Junsan Zhang; Zhimin Shao; Jie Zhu

doi:10.3788/LOP221628

Journals >Laser & Optoelectronics Progress >Volume 60 >Issue 12 >Page 1210012 > Article

Laser & Optoelectronics Progress
Vol. 60, Issue 12, 1210012 (2023)

Hyperspectral Image Classification Based on Dual-Channel Feature Enhancement

Li Zhao¹, Leiquan Wang¹, Junsan Zhang^1,*, Zhimin Shao², and Jie Zhu³

Author Affiliations

¹College of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, Shandong, China

²State Grid Shandong Electric Power Company, Jinan 250003, Shandong, China

³Department of Information Management, the National Police University for Criminal Justice, Baoding 071000, Hebei, China

show less

DOI: 10.3788/LOP221628 Cite this Article Set citation alerts

Li Zhao, Leiquan Wang, Junsan Zhang, Zhimin Shao, Jie Zhu. Hyperspectral Image Classification Based on Dual-Channel Feature Enhancement[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1210012 Copy Citation Text

show less

Fig. 1. 3D-CNN with batch normalization

Download full size

Fig. 2. Architecture diagram of CA

Download full size

Fig. 3. Multiple-branch construction

Download full size

Fig. 4. Structure of multi-branch block

Download full size

Fig. 5. DCFE network structure

Download full size

Fig. 6. Classification result diagrams of IP dataset. (a) Ground truth; (b)-(g) classification results of different methods

Download full size

Fig. 7. Classification result diagrams of UP dataset. (a) Ground-truth; (b)-(g) classification results of different methods

Download full size

Fig. 8. Classification result diagram of SV dataset. (a) Ground-truth; (b)-(g) classification results of different methods

Download full size

Fig. 9. Classification result diagram of BS dataset. (a) Ground-truth; (b)-(g) classification results of different methods

Download full size

Layer name	Kernel size	Output size
Input		（11×11×200）
Conv	（1×1×7）	（11×11×97，24）
Spectral block	（1×1×7）	（11×11×97，24）
BN-Mish-Conv	（1×1×97）	（11×11×1，24）

Table 1. Implementation of spectral-channel

Layer name	Kernel size	Output size
Input		（11×11×200）
Conv	（1×1×200）	（11×11×1，24）
Spatial block	（3×3×1）	（11×11×1，24）

Table 2. Implementation of spatial-channel

Layer name	Kernel size	Output size
Concatenate		（11×11×1，48）
Attention block		（11×11×1，48）
BN-Mish-dropout-GAP		（1×48）
Fully connected		（1×16）

Table 3. Implementation of classification module

Order	Class	Number	Training set	Verification set	Test set
Total		10249	307	307	9635
1	alfalfa	46	3	3	40
2	corn-notill	1428	42	42	1344
3	corn-mintill	830	24	24	782
4	corn	237	7	7	223
5	grass-pasture	483	14	14	455
6	grass-trees	730	21	21	688
7	grass-pasture-mowed	28	3	3	22
8	hay-windrowed	478	14	14	450
9	oats	20	3	3	14
10	soybean-notill	972	29	29	914
11	soybean-mintill	2455	73	73	2309
12	soybean-clean	593	17	17	559
13	wheat	205	6	6	193
14	woods	1265	37	37	1191
15	buildings-grass-tree-drives	386	11	11	364
16	stone-steel-towers	93	3	3	87

Table 4. Samples for each category of training, validation, and testing for IP dataset

Order	Class	Number	Training set	Verification set	Test set
Total		42776	210	210	42356
1	asphalt	6631	33	33	6465
2	meadows	18649	93	93	18463
3	gravel	2099	10	10	2079
4	corn	3064	15	15	3034
5	trees	1345	6	6	1333
6	bare soil	5029	25	25	4979
7	bitumen	1330	6	6	1318
8	self-blocking bricks	3682	18	18	3646
9	shadows	947	4	4	939

Table 5. Samples for each category of training, validation, and testing for UP dataset

Order	Class	Number	Training set	Verification set	Test set
Total		54129	263	263	53603
1	brocoli-green-weeds-1	2009	10	10	1989
2	brocoli-green-weeds-2	3726	18	18	3690
3	fallow	1976	9	9	1958
4	fallow-rough-plow	1394	6	6	1382
5	fallow-smooth	2678	13	13	2652
6	stubble	3959	19	19	3921
7	celery	3579	17	17	3545
8	grapes-untrained	11271	56	56	11159
9	soil-vinyard-develop	6203	31	31	6141
10	corn-senesced-green-weeds	3278	16	16	3246
11	lettuce-romaine-4wk	1068	5	5	1058
12	lettuce-romaine-5wk	1927	9	9	1909
13	lettuce-romaine-6wk	916	4	4	908
14	lettuce-romaine-7wk	1070	5	5	1060
15	vinyard-untrained	7268	36	36	7196
16	vinyard-vertical-trellis	1807	9	9	1789

Table 6. Samples for each category of training, validation, and testing for SV dataset

Order	Class	Number	Training set	Verification set	Test set
Total		3248	40	40	3168
1	water	270	3	3	264
2	hippo grass	101	2	2	97
3	floodplain grasses 1	251	3	3	245
4	floodplain grasses 2	215	3	3	209
5	reeds 1	269	3	3	263
6	riparian	269	3	3	263
7	fierscar 2	259	3	3	253
8	island interior	203	3	3	197
9	acacia woodlands	314	4	4	306
10	acacia shrublands	248	3	3	242
11	acacia grasslands	305	4	4	297
12	short mopane	181	2	2	177
13	mixed mopane	269	3	3	263
14	exposed soils	95	1	1	93

Table 7. Samples for each category of training, validation, and testing for BS dataset

Class	SVM	SSRN	FDSSC	DBMA	DBDA	DCFE
1 /%	24.19	67.39	97.72	61.76	87.50	100
2 /%	56.71	84.58	98.74	92.30	94.22	98.13
3 /%	65.09	92.49	97.31	97.93	98.32	94.61
4 /%	39.63	91.37	97.20	96.15	98.18	96.81
5 /%	87.33	99.04	99.53	98.00	100	97.63
6 /%	83.87	96.18	92.83	94.86	96.34	95.91
7 /%	57.20	88	100	52.94	83.33	90.90
8 /%	89.28	95.70	100	100	100	97.59
9 /%	22.58	57.14	88.88	50.00	100	100
10 /%	66.70	78.33	88.92	95.52	91.16	94.77
11 /%	62.50	95.83	99.23	95.99	97.47	96.84
12 /%	51.86	85.57	97.16	86.89	97.61	95.63
13 /%	94.79	91.86	98.90	100	97.95	100
14 /%	90.42	91.90	93.44	92.81	95.86	96.88
15 /%	62.82	90.76	95.92	90.93	93.67	96.24
16 /%	98.46	100	92.30	92.22	92.30	93.18
OA /%	69.35	90.52	96.14	93.14	96.19	96.57
AA /%	65.86	87.88	96.15	86.77	95.24	96.57
Kappa /%	64.65	89.21	95.44	92.18	95.65	96.09
Training time /s	12.23	56.06	132.43	108.67	78.96	75.41
Test time /s	1.39	3.39	5.65	7.68	6.83	7.33

Table 8. Classification results of IP dataset of 3% training samples

Class	SVM	SSRN	FDSSC	DBMA	DBDA	DCFE
1 /%	80.26	94.81	98.88	93.67	96.24	96.49
2 /%	86.94	98.50	98.82	96.34	99.23	99.26
3 /%	71.13	100	100	99.02	99.87	99.44
4 /%	96.44	100	91.74	97.43	98.20	98.78
5 /%	90.85	99.32	99.92	99.55	99.92	99.92
6 /%	77.02	93.43	99.61	98.67	98.06	99.97
7 /%	69.70	95.96	100	98.50	100	99.21
8 /%	67.30	75.87	84.02	82.48	84.11	91.19
9 /%	99.89	99.68	99.66	96.88	100	99.33
OA /%	83.07	94.85	97.02	95.06	97.11	98.15
AA /%	82.24	95.28	96.96	95.84	97.29	98.18
Kappa /%	77.07	93.17	96.04	93.40	96.17	97.54
Training time /s	5.32	12.06	32.16	29.83	21.88	20.12
Test time /s	2.19	5.21	13.22	13.52	11.25	12.10

Table 9. Classification results of UP dataset of 0.5% training samples

Class	SVM	SSRN	FDSSC	DBMA	DBDA	DCFE
1 /%	99.84	100	100	100	100	100
2 /%	98.95	100	97.20	100	97.84	100
3 /%	89.87	94.35	99.58	99.57	96.92	100
4 /%	97.30	95.63	96.91	90.26	97.71	94.33
5 /%	93.55	99.40	100	97.66	99.26	100
6 /%	99.79	100	99.74	100	99.97	99.77
7 /%	91.33	99.46	100	91.90	99.88	100
8 /%	74.73	89.14	95.15	95.62	96.53	97.32
9 /%	97.69	99.51	89.31	99.69	98.76	100
10 /%	90.01	97.75	98.17	97.38	97.70	99.28
11 /%	75.92	92.97	93.17	81.76	95.40	95.49
12 /%	95.19	99.63	98.35	95.93	99.79	100
13 /%	94.86	99.88	100	99.88	100	100
14 /%	89.26	98.04	95.92	97.62	96.00	97.78
15 /%	75.85	87.95	91.94	89.97	94.47	99.03
16 /%	99.03	100	100	100	100	100
OA /%	88.09	95.35	95.85	95.90	97.70	98.95
AA /%	91.45	97.11	97.21	96.08	98.14	98.93
Kappa /%	86.70	94.82	95.38	95.44	97.44	98.83
Training time /s	10.27	85.65	123.14	146.28	82.33	80.56
Test time /s	4.12	16.32	31.05	42.56	25.67	23.66

Table 10. Classification results of SV dataset of 0.5% training samples

Class	SVM	SSRN	FDSSC	DBMA	DBDA	DCFE
1 /%	100	100	83.95	96.33	95.97	93.26
2 /%	70.70	95.83	78.40	100	98.00	95.14
3 /%	84.10	100	95.57	100	100	100
4 /%	65.95	81.18	82.82	89.40	85.77	86.12
5 /%	82.62	84.55	100	99.45	98.96	92.30
6 /%	65.71	93.24	62.11	80.18	87.04	95.45
7 /%	78.77	94.75	98.82	84.33	100	96.93
8 /%	65.87	97.51	100	100	99.49	100
9 /%	75.18	81.74	100	100	91.04	100
10 /%	69.82	100	97.60	99.18	100	97.99
11 /%	95.49	100	99.00	99.32	100	100
12 /%	93.10	100	93.12	94.62	100	100
13 /%	76.25	100	100	100	100	100
14 /%	90.41	100	100	100	100	100
OA /%	78.63	94.27	90.80	94.87	96.39	96.83
AA /%	79.57	94.91	92.45	95.91	96.87	96.94
Kappa /%	76.87	93.79	90.03	94.45	96.09	96.57
Training time /s	1.65	10.25	22.35	20.88	18.65	19.39
Test time /s	0.41	2.01	2.37	3.02	2.11	2.04

Table 11. Classification results of BS dataset of 1.2% training samples

Algorithm	0.5%	1%	3%	5%	10%
SVM	48.53	55.95	69.35	74.74	80.55
SSRN	64.99	81.40	90.52	0.955	97.84
FDSSC	70.75	84.71	96.14	97.21	98.02
DBMA	59.33	77.64	93.14	93.75	96.91
DBDA	56.97	78.81	96.19	96.58	97.55
DCFE	74.10	86.54	96.57	97.83	98.34

Table 12. OA for different proportions of training samples in IP

Algorithm	0.1%	0.5%	1%	3%	5%
SVM	70.59	83.07	88.45	90.35	93.29
SSRN	78.32	94.85	97.11	99.43	99.69
FDSSC	88.97	97.02	97.74	99.50	99.58
DBMA	89.87	95.06	96.37	99.10	99.49
DBDA	88.01	97.11	98.40	99.07	99.33
DCFE	90.79	98.15	98.66	99.99	99.99

Table 13. OA for different proportions of training samples in UP

Algorithm	0.1%	0.5%	1%	3%	5%
SVM	78.65	88.09	89.89	91.24	92.47
SSRN	67.22	95.35	96.32	97.23	98.14
FDSSC	88.83	95.85	96.48	97.52	98.85
DBMA	92.15	95.90	96.66	97.62	98.21
DBDA	94.23	97.70	98.31	98.95	99.36
DCFE	95.70	98.95	99.25	99.81	99.98

Table 14. OA for different proportions of training samples in SV

Algorithm	0.5%	1.2%	3%	5%	10%
SVM	73.53	78.63	87.82	89.06	92.76
SSRN	84.07	94.27	95.52	98.19	99.15
FDSSC	87.98	90.80	96.33	97.24	99.46
DBMA	93.36	94.87	95.88	98.01	99.04
DBDA	96.27	96.39	97.38	98.64	99.33
DCFE	96.66	96.83	99.24	99.62	99.80

Table 15. OA for different proportions of training samples in BS

Li Zhao, Leiquan Wang, Junsan Zhang, Zhimin Shao, Jie Zhu. Hyperspectral Image Classification Based on Dual-Channel Feature Enhancement[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1210012

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享