Hyperspectral image classification combing local binary patterns and k-nearest neighbors algorithm

Jin-Ling ZHAO; Lei HU; Hao YAN; Guo-Min CHU; Yan FANG; Lin-Sheng HUANG

doi:10.11972/j.issn.1001-9014.2021.03.017

Journals >Journal of Infrared and Millimeter Waves >Volume 40 >Issue 3 >Page 400 > Article

Journal of Infrared and Millimeter Waves
Vol. 40, Issue 3, 400 (2021)

Hyperspectral image classification combing local binary patterns and k-nearest neighbors algorithm

Jin-Ling ZHAO^1、2、*, Lei HU², Hao YAN², Guo-Min CHU², Yan FANG², and Lin-Sheng HUANG^1、2、**

Author Affiliations

¹National Engineering Research Center for Agro-Ecological Big Data Analysis & Application， Anhui University， Hefei 230601，China

²School of Electronics and Information Engineering， Anhui University， Hefei 230601， China

show less

DOI: 10.11972/j.issn.1001-9014.2021.03.017 Cite this Article

Jin-Ling ZHAO, Lei HU, Hao YAN, Guo-Min CHU, Yan FANG, Lin-Sheng HUANG. Hyperspectral image classification combing local binary patterns and k-nearest neighbors algorithm[J]. Journal of Infrared and Millimeter Waves, 2021, 40(3): 400 Copy Citation Text

show less

（a） LBPs=4，r=1，（b） LBPs=8，r=1，（c） schematic diagram of LBP

Fig. 1. （a） LBP_s_=4，_r₌₁，（b） LBP_s_=8，_r₌₁，（c） schematic diagram of LBP

Download full size | View in the Article

Fig. 2. Flowchart of hyperspectral image classification based on the LBP-SSKNN

Download full size | View in the Article

Fig. 3. （a）False-color composite image（b）ground-truth classes for the Pavia University scene

Download full size | View in the Article

Fig. 4. （a）False-color composite image（b） ground truth classes for the Indian Pines scene

Download full size | View in the Article

Fig. 5. False-color composite image（a） and ground truth classes，（b） for the Salinas scene

Download full size | View in the Article

Fig. 6. Influence on classification accuracies for the number of principal components

Download full size | View in the Article

Fig. 7. Influence on classification accuracies of the Indian Pines dataset for r and s

Download full size | View in the Article

Fig. 8. Influence on classification accuracies of the Indian Pines dataset for k

Download full size | View in the Article

Fig. 9. Classification maps using the four methods for the Pavia University dataset

Download full size | View in the Article

Fig. 10. Classification maps using the four methods for the Indian Pines dataset

Download full size | View in the Article

Fig. 11. Classification maps using the four methods for the Salinas dataset

Download full size | View in the Article

Fig. 12. Classification accuracies under different training samples for the Pavia University dataset

Download full size | View in the Article

Fig. 13. Classification accuracies under different training samples for the Indian Pines dataset

Download full size | View in the Article

Fig. 14. Classification accuracies under different training samples for the Salinas dataset

Download full size | View in the Article

序号

地物类别

样本量

Asphalt

Meadows

Gravel

Trees

Painted metal sheets

Bare Soil

Bitumen

Self-Blocking Bricks

Shadows

6 631

18 649

2 099

3 064

1 345

5 029

1 330

3 682

947

Table 1. Sample information of the Pavia University dataset

View in the Article

序号

地物类别

样本量

Alfalfa

Corn-notill

Corn-mintill

Corn

Grass-pasture

Grass-trees

Grass-pasture-mowed

Hay-windrowed

Oats

Soybean-notill

Soybean-mintill

Soybean-clean

Wheat

Woods

Buildings-Grass-Trees-Drives

Stone-Steel-Towers

1 428

830

237

483

730

478

972

2 455

593

205

1 265

386

Table 2. Sample information for the Indian Pines dataset

View in the Article

序号

地物类别

样本量

Brocoli_green_weeds_1

Brocoli_green_weeds_2

Fallow

Fallow_rough_plow

Fallow_smooth

Stubble

Celery

Grapes_untrained

Soil_vinyard_develop

Corn_senesced_green_weeds

Lettuce_romaine_4wk

Lettuce_romaine_5wk

Lettuce_romaine_6wk

Lettuce_romaine_7wk

Vinyard_untrained

Vinyard_vertical_trellis

2 009

3 426

1 976

1 394

2 678

3 959

3 579

11 271

6 203

3 278

1 068

1 927

916

1 070

7 268

1 807

Table 3. Sample information for the Salinas dataset

View in the Article

主成分个数

Pavia University

Indian Pines

Salinas

贡献率/%

累计贡献率/%

贡献率/%

累计贡献率/%

贡献率/%

累计贡献率/%

58.32

36.10

4.43

0.30

0.21

0.18

0.12

0.07

58.32

94.42

98.86

99.16

99.37

99.54

99.67

99.74

68.49

23.53

1.49

0.82

0.69

0.52

0.40

0.36

68.49

92.02

93.52

94.34

95.03

95.55

95.95

96.31

74.47

23.53

1.13

0.54

0.17

0.06

0.02

0.01

74.47

98.00

99.14

99.68

99.85

99.91

99.93

99.95

Table 4. Comparison of principal component contribution for the three datasets

View in the Article

分类精度（%）	w=3	w=5	w=7	w=9	w=11	w=13	w=15
OA	97.00 ± 0.26	97.62 ± 0.36	97.72 ± 0.35	97.88 ± 0.26	97.52 ± 0.29	97.85 ± 0.28	97.71 ± 0.43
AA	95.06 ± 1.11	95.78 ± 1.62	95.59 ± 1.19	95.59 ± 1.03	96.03 ± 1.44	95.90 ± 1.41	95.80 ± 0.89
Kappa	96.58 ± 0.30	97.29 ± 0.41	97.40 ± 0.39	97.58 ± 0.29	97.17 ± 0.30	97.55 ± 0.37	97.39 ± 0.49

Table 5. Influence on classification accuracies of the Indian Pines dataset for w

View in the Article

序号	地物种类	训练样本数	测试样本数	KNN	RBF-SVM	KSOMP	LBP-SSKNN
1	Asphalt	663	5 968	80.22±0.70	91.42±0.40	87.67±0.56	99.90±0.11
2	Meadows	1 865	16 784	99.18±0.09	96.54±0.51	99.99±0.00	99.97±0.03
3	Gravel	210	1 889	64.28±2.37	71.52±2.39	89.63±3.12	99.55±0.33
4	Trees	306	2 758	79.06±0.77	92.88±0.50	94.04±1.13	92.67±1.49
5	Painted metal sheets	135	1 211	99.21±0.27	99.52±0.26	100.00±0.00	99.33±0.33
6	Bare Soil	503	4 526	40.62±1.33	75.57±1.68	85.62±2.40	100.00±0.00
7	Bitumen	133	1 197	79.39±1.51	80.07±1.55	89.49±1.09	99.89±0.15
8	Self-Blocking Bricks	368	3 314	83.84±2.01	86.06±1.24	94.78±0.77	99.48±0.62
9	Shadows	95	852	92.96±0.86	97.52±1.31	73.95±0.44	90.63±1.29
OA				84.13±0.32	90.49±0.19	94.11±0.38	99.15±0.15
AA				79.86±0.40	87.90±0.45	90.57±0.41	97.94±0.27
Kappa				78.24±0.46	87.30±0.26	92.08±0.52	98.87±0.20

Table 6. Classification accuracies using the four methods based on the Pavia University dataset

View in the Article

序号	地物种类	训练样本数	测试样本数	KNN	RBF-SVM	KSOMP	LBP-SSKNN
1	Alfalfa	5	41	24.88±10.22	17.32±14.69	86.77±7.11	93.41±4.32
2	Corn-notill	143	1 285	59.64±4.23	75.61±2.16	88.44±1.72	97.95±0.41
3	Corn-mintill	83	747	45.02±2.68	69.97±2.56	90.80±3.08	96.28±1.51
4	Corn	24	213	24.88±8.06	46.95±6.17	91.74±4.19	95.45±2.17
5	Grass-pasture	48	435	80.74±1.91	90.69±2.29	93.70±0.10	97.52±0.77
6	Grass-trees	73	657	97.35±1.57	95.25±1.44	99.06±0.39	97.84±1.58
7	Grass-pasture-mowed	3	25	66.40±9.12	21.20±15.02	47.29±9.01	94.00±4.73
8	Hay-windrowed	48	430	97.28±2.15	99.14±0.57	99.95±0.09	99.60±0.56
9	Oats	2	18	16.60±5.00	6.11±0.38	0.00±0.00	78.33±1.78
10	Soybean-notill	97	975	68.61±3.06	69.51±3.05	89.65±3.16	96.88±1.48
11	Soybean-mintill	246	2 209	73.13±1.94	86.52±1.03	96.54±0.78	98.90±0.49
12	Soybean-clean	59	534	27.08±4.68	71.89±4.92	93.61±2.39	96.72±1.51
13	Wheat	21	184	91.74±2.02	96.79±1.51	99.51±0.45	95.82±2.98
14	Woods	127	1 138	93.50±0.72	96.78±1.36	99.32±0.57	99.09±0.37
15	Buildings-Grass-Trees-Drives	39	347	15.53±1.86	52.62±4.44	87.92±7.58	98.96±0.61
16	Stone-Steel-Towers	9	84	83.93±2.08	88.81±4.03	97.63±1.30	92.74±4.90
OA				68.42±0.60	81.27±0.56	93.94±0.46	97.88±0.26
AA				59.46±1.20	67.82±0.92	85.19±0.60	95.60±1.03
Kappa				63.71±0.71	78.48±0.66	93.08±0.53	97.58±0.29

Table 7. Classification accuracies using the four methods for the Indian Pines dataset

View in the Article

序号	地物种类	训练样本数	测试样本数	KNN	RBF-SVM	KSOMP	LBP-SSKNN
1	Brocoli_green_weeds_1	40	1 969	97.17±0.48	98.60±1.15	99.91±0.12	99.81±0.32
2	Brocoli_green_weeds_2	69	3 357	98.22±0.33	99.01±0.38	99.97±0.06	99.96±0.05
3	Fallow	40	1 936	93.39±2.08	93.63±3.52	96.99±1.76	99.91±0.12
4	Fallow_rough_plow	28	1 366	98.87±0.11	98.59±0.83	99.45±0.34	93.11±2.17
5	Fallow_smooth	54	2 624	94.55±1.32	97.74±1.05	98.29±1.27	94.65±1.47
6	Stubble	79	3 880	99.52±0.09	99.44±0.23	100.00±0.00	97.32±0.95
7	Celery	72	3 507	99.21±0.08	99.35±0.25	99.05±0.42	98.56±0.75
8	Grapes_untrained	225	11 046	82.36±2.14	87.44±1.17	95.08±1.15	99.79±0.17
9	Soil_vinyard_develop	124	6 079	97.27±0.19	98.64±0.63	99.97±0.32	99.99±0.03
10	Corn_senesced_green_weeds	66	3 212	84.14±2.27	92.68±1.76	96.77±0.83	98.69±0.52
11	Lettuce_romaine_4wk	21	1 047	92.00±2.00	94.16±4.44	94.29±6.79	95.57±4.24
12	Lettuce_romaine_5wk	39	1 888	99.99±0.02	99.55±0.49	99.99±0.02	96.72±1.18
13	Lettuce_romaine_6wk	18	898	97.76±0.28	97.08±2.08	98.87±0.61	92.49±2.90
14	Lettuce_romaine_7wk	21	1 049	88.52±2.58	92.83±2.44	99.42±0.29	91.04±4.60
15	Vinyard_untrained	145	7 123	53.64±2.80	67.78±2.70	84.44±3.40	98.81±0.43
16	Vinyard_vertical_trellis	36	1 771	84.11±3.12	96.35±1.62	99.09±0.52	99.99±0.02
OA				87.02±0.26	91.43±0.39	96.23±0.40	98.46±0.15
AA				91.30±0.40	94.56±0.40	97.60±0.38	97.28±0.21
Kappa				85.51±0.29	90.44±0.44	95.81±0.45	98.29±0.17

Table 8. Classification accuracies using the four methods for the Salinas dataset

Download Citation

Save the article for my favorites

Paper Information