Hyperspectral Image Classification Combining Superpixel Principal Component Analysis Dimensionality Reduction with Extended Random Walk Probability Optimization

Dejia Hu; Yuan Huang; Bin Yang; Xinguang He

doi:10.3788/LOP220621

Laser & Optoelectronics Progress, Volume. 60, Issue 12, 1210005(2023)

Hyperspectral Image Classification Combining Superpixel Principal Component Analysis Dimensionality Reduction with Extended Random Walk Probability Optimization

Dejia Hu^1,2, Yuan Huang^1,2, Bin Yang^1,2、*, and Xinguang He^1,2

Author Affiliations

¹College of Geographic Sciences, Hunan Normal University, Changsha 410081, Hunan, China

²Hunan Key Laboratory of Geospatial Big Data Mining and Application, Changsha 410081, Hunan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(15)

Fig. 1. Hyperspectral image classification framework based on SE_SVM

Download full size

Fig. 2. False-color image (bands 34, 17, 10) and ground truth label map of Indian Pines dataset

Download full size

Fig. 3. False-color image (bands 68, 27, 19) and ground truth label map of Pavia University dataset

Download full size

Fig. 4. False-color image (bands 68, 27, 19) and ground truth label map of Salinas dataset

Download full size

Fig. 5. Line graphs of overall classification accuracy of hyperspectral dataset with the number of principal components and the number of superpixels

Download full size

Fig. 6. Classification result graphs of seven methods on Indian Pines dataset. (a) Ground truth label map; (b) SVM; (c) PCA_SVM; (d) SPCA_SVM; (e) ERW_SVM; (f) 3DCNN; (g) SSRN; (h) SE_SVM

Download full size

Fig. 7. Classification result graphs of seven methods on the Pavia University dataset. (a) Ground truth label map; (b) SVM; (c) PCA_SVM;(d) SPCA_SVM; (e) ERW_SVM; (f) 3DCNN; (g) SSRN; (h) SE_SVM

Download full size

Fig. 8. Classification result graphs of seven methods on the Salinas dataset. (a) Ground truth label map; (b) SVM; (c) PCA_SVM; (d) SPCA_SVM; (e) ERW_SVM; (f) 3DCNN; (g) SSRN; (h) SE_SVM

Download full size

Table 1. Data partitioning of Indian Pines dataset

View table

Table 1. Data partitioning of Indian Pines dataset

Label	Class	Labeled sample	Training	Validation	Test
Total		10249	512	511	9226
1	Alfalfa	46	3	2	41
2	Corn-notill	1428	71	71	1286
3	Corn-mintill	830	43	43	744
4	Corn	237	12	12	213
5	Grass-pasture	483	24	24	435
6	Grass-trees	730	36	36	658
7	Grass-pasture-mowed	28	2	1	25
8	Hay-windrowed	478	24	24	430
9	Oats	20	1	1	18
10	Soybean-notill	972	48	48	876
11	Soybean-mintill	2455	122	122	2211
12	Soybean-clean	593	30	30	533
13	Wheat	205	10	10	185
14	Woods	1265	63	63	1139
15	Buildings-Grass-Trees-Drives	386	19	19	348
16	Stone-Steel-Towers	93	4	5	84

Table 2. Data partitioning of Pavia University dataset

View table

Table 2. Data partitioning of Pavia University dataset

Label	Class	Labeled sample	Training	Validation	Test
Total		42776	426	426	41924
1	Asphalt	6631	66	66	6499
2	Meadows	18649	186	186	18277
3	Gravel	2099	21	21	2057
4	Trees	3064	31	31	3002
5	Painted metal sheets	1345	13	13	1319
6	Bare Soil	5029	50	50	4929
7	Bitumen	1330	13	13	1304
8	Self-Blocking Bricks	3682	37	37	3608
9	Shadows	947	9	9	929

Table 3. Data partitioning of Salinas dataset

View table

Table 3. Data partitioning of Salinas dataset

Label	Class	Labeled sample	Training	Validation	Test
Total		54129	542	542	53045
1	Brocoli_green_weeds_1	2009	20	20	1969
2	Brocoli_green_weeds_2	3726	37	37	3652
3	Fallow	1976	20	20	1936
4	Fallow_rough_plow	1394	14	14	1366
5	Fallow_smooth	2678	27	27	2624
6	Stubble	3959	40	40	3879
7	Celery	3579	36	36	3507
8	Grapes_untrained	11271	112	112	11047
9	Soil_vinyard_develop	6203	62	62	6079
10	Corn_senesced_green_weeds	3278	33	33	3212
11	Lettuce_romaine_4wk	1068	11	11	1046
12	Lettuce_romaine_5wk	1927	19	19	1889
13	Lettuce_romaine_6wk	916	9	9	898
14	Lettuce_romaine_7wk	1070	11	11	1048
15	Vinyard_untrained	7268	73	73	7122
16	Vinyard_vertical_trellis	1807	18	18	1771

Table 4. Classification accuracy achieved by seven different methods on Indian Pines dataset

View table

Table 4. Classification accuracy achieved by seven different methods on Indian Pines dataset

Label	SVM	PCA_SVM	SPCA_SVM	ERW_SVM	3DCNN	SSRN	SE_SVM
1	0.00	16.73	0.00	100.00	26.83	36.59	100.00
2	61.95	52.39	93.23	95.84	81.17	96.81	97.56
3	69.03	62.02	92.67	98.40	79.65	98.13	98.42
4	57.75	50.24	91.86	99.13	67.14	69.48	94.51
5	85.68	75.75	97.78	98.54	86.21	97.70	99.20
6	83.66	78.86	99.69	95.76	97.26	99.70	98.72
7	0.00	0.00	0.00	99.50	36.00	0.00	98.80
8	85.07	90.09	100.00	99.26	99.77	100.00	100.00
9	0.00	0.00	0.00	100.00	0.00	0.00	98.00
10	74.08	64.40	94.42	97.57	91.31	97.14	98.57
11	67.99	58.10	91.32	96.72	84.48	98.51	98.05
12	59.86	50.14	87.28	95.70	66.48	92.70	96.81
13	91.84	91.34	100.00	99.74	98.38	100.00	99.79
14	89.88	83.89	99.30	95.31	98.42	100.00	98.94
15	59.72	55.80	98.52	98.64	72.62	94.24	99.83
16	98.29	97.61	97.66	96.21	25.00	38.10	98.90
OA	73.28	65.99	94.38	96.84	85.00	95.99	98.29
AA	61.55	57.96	77.73	97.90	69.42	76.19	98.51
Kappa	69.12	60.21	93.57	96.40	82.89	95.42	98.05

Table 5. Classification accuracy achieved by seven different methods on Pavia University dataset

View table

Table 5. Classification accuracy achieved by seven different methods on Pavia University dataset

Label	SVM	PCA_SVM	SPCA_SVM	ERW_SVM	3DCNN	SSRN	SE_SVM
1	84.80	80.70	78.26	94.61	89.00	99.45	94.37
2	91.89	89.08	96.45	96.65	97.76	98.73	97.34
3	77.97	73.95	86.21	99.84	86.78	88.24	98.69
4	94.15	89.46	96.85	99.54	89.48	90.41	98.57
5	98.94	100.00	99.74	99.49	89.38	99.93	99.92
6	85.35	84.04	95.87	97.80	78.86	99.19	97.43
7	81.84	66.80	69.15	100.00	81.20	95.09	99.72
8	76.25	68.84	80.82	97.57	82.32	89.95	98.47
9	100.00	99.99	99.08	99.80	94.40	98.72	99.73
OA	88.21	84.83	91.08	97.06	90.86	96.96	97.29
AA	87.91	83.65	89.16	98.37	87.68	95.52	98.25
Kappa	84.19	79.54	88.07	96.07	87.82	96.02	96.38

Table 6. Classification accuracy achieved by seven different methods on Salinas dataset

View table

Table 6. Classification accuracy achieved by seven different methods on Salinas dataset

Label	SVM	PCA_SVM	SPCA_SVM	ERW_SVM	3DCNN	SSRN	SE_SVM
1	99.52	98.98	100.00	100.00	97.16	99.44	99.99
2	99.10	98.53	100.00	100.00	100.00	100.00	100.00
3	93.51	90.76	100.00	99.30	99.48	100.00	99.86
4	97.76	98.66	96.30	96.08	99.56	99.93	98.52
5	98.05	93.56	97.79	99.99	95.01	97.48	99.97
6	99.95	99.82	100.00	100.00	99.90	100.00	99.97
7	98.38	97.83	99.99	99.85	97.38	99.91	99.95
8	74.15	72.72	98.98	97.36	92.80	85.79	99.98
9	98.69	97.75	99.03	99.98	99.87	100.00	100.00
10	85.11	88.60	95.35	98.70	94.43	97.54	98.42
11	91.71	90.66	92.89	99.74	91.98	95.32	100.00
12	95.74	97.67	93.40	100.00	98.73	100.00	100.00
13	96.49	96.13	100.00	99.24	98.55	97.88	99.39
14	97.06	97.15	97.04	99.32	98.28	98.76	95.00
15	73.66	76.39	99.99	99.12	70.67	97.29	99.87
16	98.34	99.69	100.00	100.00	91.81	97.97	100.00
OA	88.85	88.52	98.65	99.06	93.11	96.15	99.72
AA	93.58	93.43	98.17	99.29	95.35	97.95	99.43
Kappa	87.55	87.17	98.50	98.96	92.31	95.72	99.69

Table 7. Running time of five methods on the Indian Pines, Pavia University, and Salinas datasets
View table
Table 7. Running time of five methods on the Indian Pines, Pavia University, and Salinas datasets
Dataset SVM PCA_SVM SPCA_SVM ERW_SVM SE_SVM
Indian Pines 42.87 5.74 10.38 80.82 15.49
Pavia University 15.16 10.2 25.12 35.48 16.71
Salinas 40.63 12.83 26.63 85.18 19.58

Tools

Get Citation

Copy Citation Text

Dejia Hu, Yuan Huang, Bin Yang, Xinguang He. Hyperspectral Image Classification Combining Superpixel Principal Component Analysis Dimensionality Reduction with Extended Random Walk Probability Optimization[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1210005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Jan. 25, 2022

Accepted: Jun. 14, 2022

Published Online: Jun. 5, 2023

The Author Email: Yang Bin (yangbin@hunnu.edu.cn)

DOI:10.3788/LOP220621

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Data partitioning of Indian Pines dataset

Table 1. Data partitioning of Indian Pines dataset

Table 2. Data partitioning of Pavia University dataset

Table 2. Data partitioning of Pavia University dataset

Table 3. Data partitioning of Salinas dataset

Table 3. Data partitioning of Salinas dataset

Table 4. Classification accuracy achieved by seven different methods on Indian Pines dataset

Table 4. Classification accuracy achieved by seven different methods on Indian Pines dataset

Table 5. Classification accuracy achieved by seven different methods on Pavia University dataset

Table 5. Classification accuracy achieved by seven different methods on Pavia University dataset

Table 6. Classification accuracy achieved by seven different methods on Salinas dataset

Table 6. Classification accuracy achieved by seven different methods on Salinas dataset

Table 7. Running time of five methods on the Indian Pines, Pavia University, and Salinas datasets

Table 7. Running time of five methods on the Indian Pines, Pavia University, and Salinas datasets