Hyperspectral Image Classification Based on Local Gaussian Mixture Feature Extraction

Dan Li; Fanqiang Kong; Deyan Zhu

doi:10.3788/AOS202141.0610001

Acta Optica Sinica, Volume. 41, Issue 6, 0610001(2021)

Hyperspectral Image Classification Based on Local Gaussian Mixture Feature Extraction

Dan Li^1,2、*, Fanqiang Kong², and Deyan Zhu^1,2

¹Key Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu 210016, China

²College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu 210016, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(20)

Fig. 1. Denoising results of MNF method in Indian Pines dataset. (a) 1^st band of Indian Pines dataset; (b) 5^th band of Indian Pines dataset; (c) 1^st band after dimensionality reduction

Download full size

Fig. 2. Flowchart of proposed method

Download full size

Fig. 3. Indian Pines dataset. (a) False colour image; (b) correct classification diagram; (c) category name

Download full size

Fig. 4. Pavia University dataset. (a) False colour image; (b) correct classification diagram; (c) category name

Download full size

Fig. 5. Salinas dataset. (a) False colour image; (b) correct classification diagram; (c) category name

Download full size

Fig. 6. OA curves of three groups of hyperspectral images at different r values

Download full size

Fig. 7. OA curves of three groups of hyperspectral images at different Q values

Download full size

Fig. 8. OA curves of three groups of hyperspectral images at different m values

Download full size

Fig. 9. Classification results of different algorithms on Indian Pines dataset. (a) SVM; (b) LBP; (c) SC-MK; (d) MFASRC; (e) MFKSRC; (f) LCMR; (g) Semi-Sup; (h) LGMFEC

Download full size

Fig. 10. Classification results of different algorithms on Pavia University dataset. (a) SVM; (b) LBP; (c) SC-MK; (d) MFASRC; (e) MFKSRC; (f) LCMR; (g) Semi-Sup; (h) LGMFEC

Download full size

Fig. 11. Classification results of different algorithms on Salinas dataset. (a) SVM; (b) LBP; (c) SC-MK; (d) MFASRC; (e) MFKSRC; (f) LCMR; (g) Semi-SUP; (h) LGMFEC

Download full size

Fig. 12. Classification accuracy of different algorithms in Indian Pines dataset

Download full size

Fig. 13. Classification accuracy of different algorithms in Pavia University dataset

Download full size

Fig. 14. Classification accuracy of different algorithms in Salinas dataset

Download full size

Table 1. Number of different samples in Indian Pines dataset

View table

Table 1. Number of different samples in Indian Pines dataset

Class	Name	Training	Test
1	Alfalfa	5	41
2	Corn-no till	5	1423
3	Corn-min till	5	825
4	Corn	5	232
5	Grass/pasture	5	478
6	Grass/tree	5	725
7	Grass/pasture-mowed	5	24
8	Hay-windrowed	5	473
9	Oat	5	15
10	Soybean-no till	5	967
11	Soybean-min till	5	2450
12	Soybean-clean till	5	588
13	Wheat	5	200
14	Wood	5	1260
15	Bldg-grass-tree-drive	5	381
16	Stone-steel	5	88
Total		80	10170

Table 2. Number of different samples in Pavia University dataset
View table
Table 2. Number of different samples in Pavia University dataset
Class Name Training Test
1 Asphalt 10 6621
2 Meadow 10 18639
3 Gravel 10 2089
4 Tree 10 3054
5 Metrl sheet 10 1335
6 Grass/tree 10 5019
7 Bitumen 10 1320
8 Brick 10 3672
9 Shadow 10 937
Total 90 42686

Table 3. Number of different samples in Salinas dataset

View table

Table 3. Number of different samples in Salinas dataset

Class	Name	Training	Test
1	Broccoli-green-weed 1	5	2004
2	Broccoli-green-weed 2	5	3721
3	Fallow	5	1971
4	Fallow-rough-plow	5	1386
5	Fallow-smooth	5	2673
6	Stubble	5	3954
7	Celery	5	3574
8	Grapes-untrained	5	11266
9	Soil-vinyary-develop	5	6198
10	Corn-senesced-green-weed	5	3273
11	Lettuce-romaine-4wk	5	1063
12	Lettuce-romaine-5wk	5	1922
13	Lettuce-romaine-6wk	5	911
14	Lettuce-romaine-7wk	5	1065
15	Vineyard-untrained	5	7263
16	Vineyard-vertical-trellis	5	1802
Total		80	54046

Table 4. Classification results of different algorithms on Indian Pines dataset

View table

Table 4. Classification results of different algorithms on Indian Pines dataset

Class	SVM	LBP	SC-MK	MFASRC	MFKSRC	LCMR	Semi-Sup	LGMFEC
1	85.85	99.77	98.05	98.05	100.00	99.76	100.00	100.00
2	44.67	55.46	54.98	49.66	53.97	63.33	67.50	68.51
3	52.00	56.53	53.54	64.78	58.23	51.68	80.48	82.25
4	60.52	86.75	86.38	69.48	86.24	87.93	97.84	82.89
5	77.64	72.96	62.47	77.59	76.79	79.02	80.15	80.46
6	74.08	72.01	90.91	78.19	69.88	83.60	78.34	90.83
7	91.30	99.20	98.70	100.00	99.60	100.00	100.00	100.00
8	71.06	98.34	100.00	99.09	98.78	98.84	99.79	99.81
9	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
10	51.33	66.90	66.50	68.54	67.42	69.45	73.73	72.17
11	40.84	60.86	57.96	56.77	66.05	59.88	61.35	64.29
12	53.37	51.27	62.98	65.48	61.56	71.14	74.97	78.42
13	98.00	97.62	96.35	99.50	99.50	99.15	99.15	99.85
14	81.34	93.90	74.45	87.83	91.79	95.27	88.49	91.21
15	45.91	78.75	79.45	93.73	74.60	79.74	98.43	86.51
16	90.68	99.00	97.95	98.41	97.56	96.59	99.50	99.77
OA /%	57.10	69.06	67.72	69.57	71.09	72.83	76.15	78.08
AA /%	69.91	79.58	80.04	81.69	81.38	83.46	87.25	87.28
κ	52.12	64.97	62.40	65.81	67.54	69.43	73.68	75.36

Table 5. Classification results of different algorithms on Pavia University dataset

View table

Table 5. Classification results of different algorithms on Pavia University dataset

Class	SVM	LBP	SC-MK	MFASRC	MFKSRC	LCMR	Semi-Sup	LGMFEC
1	60.85	65.23	87.20	91.52	68.82	82.36	64.93	84.95
2	62.72	59.37	77.68	70.94	72.28	86.13	85.33	90.33
3	65.38	88.33	83.00	92.34	94.27	87.73	85.43	95.49
4	88.63	86.47	90.24	93.71	88.34	96.32	90.65	93.97
5	99.75	100.00	99.75	99.93	99.96	99.10	99.93	99.39
6	80.02	88.20	86.54	89.45	86.42	91.08	88.47	93.93
7	81.89	95.83	92.32	99.90	98.17	95.60	99.09	98.83
8	69.56	89.67	85.45	96.12	89.59	83.19	88.55	87.54
9	92.33	95.75	91.25	99.32	95.15	94.91	78.66	84.83
OA /%	67.41	72.95	83.47	83.82	79.29	87.58	83.99	90.62
AA /%	77.85	85.10	88.16	92.58	88.11	90.71	86.78	92.14
κ	59.42	66.22	81.14	79.41	73.99	83.96	80.07	87.83

Table 6. Classification results of different algorithms in Salinas dataset

View table

Table 6. Classification results of different algorithms in Salinas dataset

Class	SVM	LBP	SC-MK	MFASRC	MFKSRC	LCMR	Semi-Sup	LGMFEC
1	98.92	99.82	98.07	99.83	100.00	99.97	100.00	100.00
2	94.30	96.22	99.48	98.01	96.88	75.72	100.00	100.00
3	89.98	78.84	99.43	90.33	77.84	97.98	93.30	99.87
4	93.81	94.67	96.18	99.79	99.63	96.18	98.27	99.29
5	88.57	93.91	93.45	98.27	97.88	90.66	98.68	98.83
6	99.65	99.72	99.75	99.73	99.72	99.22	99.22	99.82
7	98.03	95.32	99.47	98.64	96.01	96.16	99.92	99.92
8	53.35	67.25	59.65	70.94	58.26	84.35	59.80	83.31
9	97.47	86.84	99.90	99.32	86.86	83.62	99.95	99.99
10	68.30	72.01	88.49	82.64	89.92	91.23	95.08	96.43
11	94.27	93.56	97.66	97.93	96.73	99.97	95.48	96.63
12	92.96	83.61	88.20	99.89	93.40	96.75	97.50	99.94
13	98.96	89.18	92.28	97.19	93.49	97.46	95.28	96.20
14	91.48	95.85	88.25	94.88	95.97	95.75	94.93	93.30
15	67.93	74.36	80.24	74.71	79.67	83.80	99.55	86.58
16	91.69	73.65	89.88	85.80	87.41	92.81	99.00	99.01
OA /%	81.26	82.64	86.46	88.05	84.12	89.05	90.46	94.10
AA /%	88.75	87.18	91.92	92.99	90.60	92.60	95.38	96.82
κ	79.22	81.02	85.40	86.74	82.44	87.84	89.67	93.44

Tools

Get Citation

Copy Citation Text

Dan Li, Fanqiang Kong, Deyan Zhu. Hyperspectral Image Classification Based on Local Gaussian Mixture Feature Extraction[J]. Acta Optica Sinica, 2021, 41(6): 0610001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites