Hyperspectral Image Classification Based on Enhanced Dynamic-Graph-Convolutional Feature Extraction

Fig. 9. Classification visualization comparison of Indian Pines dataset. (a) False colour; (b) label diagram; (c) SVM; (d) HybridSN; (e) S²RGAnet; (f) EGNN; (g) SSPGAT; (h) DCGA

Download full size

Fig. 10. Classification visualization comparison of WHU-Hi-HongHu dataset. (a) False colour; (b) label diagram; (c) SVM; (d) HybridSN; (e) S²RGAnet; (f) EGNN; (g) SSPGAT; (h) DCGA

Download full size

Fig. 11. Classification visualization comparison of WHU-Hi-HanChuan dataset. (a) False colour; (b) label diagram; (c) SVM; (d) HybridSN; (e) S²RGAnet; (f) EGNN; (g) SSPGAT; (h) DCGA

Download full size

Fig. 12. Comparison of small sample classification performance of different algorithms. (a) Indian Pines; (b) WHU-Hi-HanChuan

Download full size

Table 1. Comparative experiments on the Indian Pines dataset

View table

Table 1. Comparative experiments on the Indian Pines dataset

Class	SVM	HybridSN	S²RGAnet	EGNN	SSPGAT	DCGA
1	35.64	88.89	97.50	100.00	100.00	100.00
2	62.49	86.45	95.73	95.49	92.30	99.25
3	68.18	92.97	95.40	99.43	94.37	99.84
4	52.33	95.38	98.28	91.70	90.82	98.35
5	86.40	92.91	98.90	98.81	99.76	100.00
6	85.47	96.54	98.68	98.59	96.80	98.40
7	75.75	61.11	82.10	64.52	62.50	100.00
8	89.27	96.35	95.88	97.70	98.65	100.00
9	59.78	41.18	77.28	76.19	62.50	100.00
10	68.84	90.84	94.58	93.20	95.44	93.79
11	70.44	94.72	98.32	98.76	95.78	99.09
12	61.36	88.56	98.09	96.01	84.62	98.04
13	87.06	97.73	100.00	100.00	96.43	100.00
14	89.59	97.67	97.08	99.02	98.96	99.59
15	68.52	96.51	95.96	98.56	95.33	99.66
16	98.64	84.71	97.48	87.06	79.12	91.78
OA	73.81	92.93	97.08	97.19	94.76	98.43
AA	72.48	85.70	95.55	95.40	89.42	98.26
Kappa	69.88	91.92	96.67	96.80	94.02	98.14

Table 2. Comparative experiments on the WHU-Hi-HongHu dataset

View table

Table 2. Comparative experiments on the WHU-Hi-HongHu dataset

Class	SVM	HybridSN	S²RGAnet	EGNN	SSPGAT	DCGA
1	90.80	99.50	98.57	97.97	98.99	99.62
2	82.44	83.06	96.11	94.85	92.24	96.02
3	79.32	97.75	95.97	97.73	94.01	97.67
4	98.51	99.56	99.77	99.69	99.51	99.81
5	87.28	96.80	98.89	98.11	98.04	98.85
6	95.51	98.48	96.99	99.48	98.48	99.34
7	84.23	93.90	93.66	95.86	95.75	97.23
8	40.31	92.51	91.71	95.22	88.52	97.31
9	90.59	97.63	99.22	98.94	99.61	99.52
10	69.13	91.23	95.74	98.21	93.31	96.96
11	76.91	93.41	96.01	97.40	94.67	98.20
12	53.23	90.19	94.21	92.23	87.22	95.24
13	72.83	91.59	97.40	95.65	94.91	97.81
14	81.83	90.35	97.92	96.39	94.69	95.53
15	65.13	87.12	91.41	94.19	97.90	95.12
16	89.67	98.92	98.85	98.21	96.94	99.18
17	67.34	98.10	97.06	98.84	95.13	99.01
18	78.54	91.11	96.62	98.27	95.29	99.20
19	85.82	91.52	88.30	96.94	91.55	95.88
20	69.98	94.65	93.74	98.52	94.66	94.92
21	40.27	92.07	92.04	89.58	85.89	96.78
22	89.11	98.21	98.09	97.16	97.81	99.76
OA	85.68	96.94	97.66	98.33	97.24	98.76
AA	72.03	93.22	93.67	96.67	92.75	97.25
Kappa	84.65	96.13	97.04	97.89	96.51	98.43

Table 3. Comparative experiments on the WHU-Hi-HanChuan dataset

View table

Table 3. Comparative experiments on the WHU-Hi-HanChuan dataset

Class	SVM	HybridSN	S²RGAnet	EGNN	SSPGAT	DCGA
1	85.55	96.87	97.06	96.25	97.57	97.48
2	57.98	93.14	96.58	94.92	94.43	98.26
3	60.31	95.82	95.79	98.13	99.80	96.43
4	83.56	94.99	94.41	97.59	90.84	96.96
5	12.46	95.54	93.66	98.15	93.18	85.51
6	20.76	77.79	79.41	89.16	84.77	78.66
7	65.24	83.81	93.73	92.46	89.46	93.22
8	67.21	89.77	97.05	92.14	94.47	96.38
9	54.68	88.14	90.49	89.25	91.29	94.67
10	94.32	95.96	92.91	94.32	96.14	97.71
11	86.10	91.55	97.83	98.13	98.01	95.71
12	43.26	77.48	83.95	90.05	90.51	91.04
13	46.59	81.03	82.78	89.21	77.64	88.07
14	79.31	97.19	94.86	97.06	96.86	96.94
15	95.95	76.50	94.26	98.92	60.44	95.16
16	97.99	99.59	99.89	99.51	99.83	99.90
OA	79.84	94.49	96.04	96.23	95.91	96.93
AA	65.70	88.66	92.75	89.95	90.61	94.34
Kappa	76.33	93.55	95.37	95.58	95.22	96.40

Table 4. Comparison of DCGA ablation in three datasets
View table
Table 4. Comparison of DCGA ablation in three datasets
Parameter Indian Pines WHU-Hi-HongHu WHU-Hi-HanChuan
DGCN DSConv DGCN DSConv DGCN DSConv
OA 92.93 93.84 91.98 93.87 90.12 92.34
AA 90.14 91.37 88.48 85.53 86.34 87.67
Kappa 91.45 92.95 89.74 92.24 88.47 91.01

Tools

Get Citation

Copy Citation Text

Tie Li, Qiaoyu Gao, Wenxu Li. Hyperspectral Image Classification Based on Enhanced Dynamic-Graph-Convolutional Feature Extraction[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1837013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites