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

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

Jin-Ling ZHAO1,2、*, Lei HU2, Hao YAN2, Guo-Min CHU2, Yan FANG2, and Lin-Sheng HUANG1,2、**
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
    Figures & Tables(22)
    (a) LBPs=4,r=1,(b) LBPs=8,r=1,(c) schematic diagram of LBP
    Flowchart of hyperspectral image classification based on the LBP-SSKNN
    (a)False-color composite image(b)ground-truth classes for the Pavia University scene
    (a)False-color composite image(b) ground truth classes for the Indian Pines scene
    False-color composite image(a) and ground truth classes,(b) for the Salinas scene
    Influence on classification accuracies for the number of principal components
    Influence on classification accuracies of the Indian Pines dataset for r and s
    Influence on classification accuracies of the Indian Pines dataset for k
    Classification maps using the four methods for the Pavia University dataset
    Classification maps using the four methods for the Indian Pines dataset
    Classification maps using the four methods for the Salinas dataset
    Classification accuracies under different training samples for the Pavia University dataset
    Classification accuracies under different training samples for the Indian Pines dataset
    Classification accuracies under different training samples for the Salinas dataset
    • Table 1. Sample information of the Pavia University dataset

      View table
      View in Article

      Table 1. Sample information of the Pavia University dataset

      序号地物类别样本量

      1

      2

      3

      4

      5

      6

      7

      8

      9

      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 2. Sample information for the Indian Pines dataset

      View table
      View in Article

      Table 2. Sample information for the Indian Pines dataset

      序号地物类别样本量

      1

      2

      3

      4

      5

      6

      7

      8

      9

      10

      11

      12

      13

      14

      15

      16

      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

      46

      1 428

      830

      237

      483

      730

      28

      478

      20

      972

      2 455

      593

      205

      1 265

      386

      93

    • Table 3. Sample information for the Salinas dataset

      View table
      View in Article

      Table 3. Sample information for the Salinas dataset

      序号地物类别样本量

      1

      2

      3

      4

      5

      6

      7

      8

      9

      10

      11

      12

      13

      14

      15

      16

      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 4. Comparison of principal component contribution for the three datasets

      View table
      View in Article

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

      主成分个数Pavia UniversityIndian PinesSalinas
      贡献率/%累计贡献率/%贡献率/%累计贡献率/%贡献率/%累计贡献率/%

      1

      2

      3

      4

      5

      6

      7

      8

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

      View table
      View in Article

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

      分类精度(%)w=3w=5w=7w=9w=11w=13w=15
      OA97.00 ± 0.2697.62 ± 0.3697.72 ± 0.3597.88 ± 0.2697.52 ± 0.2997.85 ± 0.2897.71 ± 0.43
      AA95.06 ± 1.1195.78 ± 1.6295.59 ± 1.1995.59 ± 1.0396.03 ± 1.4495.90 ± 1.4195.80 ± 0.89
      Kappa96.58 ± 0.3097.29 ± 0.4197.40 ± 0.3997.58 ± 0.2997.17 ± 0.3097.55 ± 0.3797.39 ± 0.49
    • Table 6. Classification accuracies using the four methods based on the Pavia University dataset

      View table
      View in Article

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

      序号地物种类训练样本数测试样本数KNNRBF-SVMKSOMPLBP-SSKNN
      1Asphalt6635 96880.22±0.7091.42±0.4087.67±0.5699.90±0.11
      2Meadows1 86516 78499.18±0.0996.54±0.5199.99±0.0099.97±0.03
      3Gravel2101 88964.28±2.3771.52±2.3989.63±3.1299.55±0.33
      4Trees3062 75879.06±0.7792.88±0.5094.04±1.1392.67±1.49
      5Painted metal sheets1351 21199.21±0.2799.52±0.26100.00±0.0099.33±0.33
      6Bare Soil5034 52640.62±1.3375.57±1.6885.62±2.40100.00±0.00
      7Bitumen1331 19779.39±1.5180.07±1.5589.49±1.0999.89±0.15
      8Self-Blocking Bricks3683 31483.84±2.0186.06±1.2494.78±0.7799.48±0.62
      9Shadows9585292.96±0.8697.52±1.3173.95±0.4490.63±1.29
      OA84.13±0.3290.49±0.1994.11±0.3899.15±0.15
      AA79.86±0.4087.90±0.4590.57±0.4197.94±0.27
      Kappa78.24±0.4687.30±0.2692.08±0.5298.87±0.20
    • Table 7. Classification accuracies using the four methods for the Indian Pines dataset

      View table
      View in Article

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

      序号地物种类训练样本数测试样本数KNNRBF-SVMKSOMPLBP-SSKNN
      1Alfalfa54124.88±10.2217.32±14.6986.77±7.1193.41±4.32
      2Corn-notill1431 28559.64±4.2375.61±2.1688.44±1.7297.95±0.41
      3Corn-mintill8374745.02±2.6869.97±2.5690.80±3.0896.28±1.51
      4Corn2421324.88±8.0646.95±6.1791.74±4.1995.45±2.17
      5Grass-pasture4843580.74±1.9190.69±2.2993.70±0.1097.52±0.77
      6Grass-trees7365797.35±1.5795.25±1.4499.06±0.3997.84±1.58
      7Grass-pasture-mowed32566.40±9.1221.20±15.0247.29±9.0194.00±4.73
      8Hay-windrowed4843097.28±2.1599.14±0.5799.95±0.0999.60±0.56
      9Oats21816.60±5.006.11±0.380.00±0.0078.33±1.78
      10Soybean-notill9797568.61±3.0669.51±3.0589.65±3.1696.88±1.48
      11Soybean-mintill2462 20973.13±1.9486.52±1.0396.54±0.7898.90±0.49
      12Soybean-clean5953427.08±4.6871.89±4.9293.61±2.3996.72±1.51
      13Wheat2118491.74±2.0296.79±1.5199.51±0.4595.82±2.98
      14Woods1271 13893.50±0.7296.78±1.3699.32±0.5799.09±0.37
      15Buildings-Grass-Trees-Drives3934715.53±1.8652.62±4.4487.92±7.5898.96±0.61
      16Stone-Steel-Towers98483.93±2.0888.81±4.0397.63±1.3092.74±4.90
      OA68.42±0.6081.27±0.5693.94±0.4697.88±0.26
      AA59.46±1.2067.82±0.9285.19±0.6095.60±1.03
      Kappa63.71±0.7178.48±0.6693.08±0.5397.58±0.29
    • Table 8. Classification accuracies using the four methods for the Salinas dataset

      View table
      View in Article

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

      序号地物种类训练样本数测试样本数KNNRBF-SVMKSOMPLBP-SSKNN
      1Brocoli_green_weeds_1401 96997.17±0.4898.60±1.1599.91±0.1299.81±0.32
      2Brocoli_green_weeds_2693 35798.22±0.3399.01±0.3899.97±0.0699.96±0.05
      3Fallow401 93693.39±2.0893.63±3.5296.99±1.7699.91±0.12
      4Fallow_rough_plow281 36698.87±0.1198.59±0.8399.45±0.3493.11±2.17
      5Fallow_smooth542 62494.55±1.3297.74±1.0598.29±1.2794.65±1.47
      6Stubble793 88099.52±0.0999.44±0.23100.00±0.0097.32±0.95
      7Celery723 50799.21±0.0899.35±0.2599.05±0.4298.56±0.75
      8Grapes_untrained22511 04682.36±2.1487.44±1.1795.08±1.1599.79±0.17
      9Soil_vinyard_develop1246 07997.27±0.1998.64±0.6399.97±0.3299.99±0.03
      10Corn_senesced_green_weeds663 21284.14±2.2792.68±1.7696.77±0.8398.69±0.52
      11Lettuce_romaine_4wk211 04792.00±2.0094.16±4.4494.29±6.7995.57±4.24
      12Lettuce_romaine_5wk391 88899.99±0.0299.55±0.4999.99±0.0296.72±1.18
      13Lettuce_romaine_6wk1889897.76±0.2897.08±2.0898.87±0.6192.49±2.90
      14Lettuce_romaine_7wk211 04988.52±2.5892.83±2.4499.42±0.2991.04±4.60
      15Vinyard_untrained1457 12353.64±2.8067.78±2.7084.44±3.4098.81±0.43
      16Vinyard_vertical_trellis361 77184.11±3.1296.35±1.6299.09±0.5299.99±0.02
      OA87.02±0.2691.43±0.3996.23±0.4098.46±0.15
      AA91.30±0.4094.56±0.4097.60±0.3897.28±0.21
      Kappa85.51±0.2990.44±0.4495.81±0.4598.29±0.17
    Tools

    Get Citation

    Copy Citation Text

    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

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Jun. 29, 2020

    Accepted: --

    Published Online: Sep. 9, 2021

    The Author Email: Jin-Ling ZHAO (zhaojl@ahu.edu.cn), Lin-Sheng HUANG (linsheng0808@163.com)

    DOI:10.11972/j.issn.1001-9014.2021.03.017

    Topics