Remote Sensing Technology and Application, Volume. 39, Issue 4, 880(2024)

Research on UAV Hyperspectral of Tree Species Classification based on Machine Learning Algorithms and Spatial Resolution Adjustment

Xiangshan ZHOU, Wunian YANG, Ke LUO, Hongyi PIAO, Tao ZHOU, Jie ZHOU, and Xiaolu TANG
Author Affiliations
  • POWERCHINA Chengdu Engineering Corporation Limited, Chengdu611100, China
  • show less
    Figures & Tables(13)
    Overview of Chengdu botanical garden and distribution of sample points
    Technical route of UAV hyperspectral tree species classification method based on machine learning algorithm
    Removal of spectral envelope lines for 20 tree species
    First derivative of spectra for 20 tree species
    Classification accuracy trends of random forests and support vector machine algorithms based on different spatial resolutions of images for different tree species in the research area
    • Table 1. List of 20 tree types sampl in the study area

      View table
      View in Article

      Table 1. List of 20 tree types sampl in the study area

      树种样本数(棵)树种样本数(棵)树种样本数(棵)树种样本数(棵)
      桉树30红豆杉38木芙蓉32雪松22
      灯台36梨树32桑树31银桦27
      枫香31栎树40水杉29银木22
      桂花41栾树43喜树24皂荚32
      含笑3032香椿30樟树114
    • Table 2. Original band information of hyperspectral image

      View table
      View in Article

      Table 2. Original band information of hyperspectral image

      波段波段范围/nm空间分辨率/m
      B1~B12380~430(紫)0.12
      B13~B24430~470(蓝)0.12
      B25~B33470~500(青)0.12
      B34~B51500~560(绿)0.12
      B52~B60560~590(黄)0.12
      B61~B69590~620(橙)0.12
      B70~B110620~760(红)0.12
      B111~B176760~1 000(近红外)0.12
    • Table 3. Calculation of hyperspectral image vegetation index

      View table
      View in Article

      Table 3. Calculation of hyperspectral image vegetation index

      序号名称计算公式参考文献
      1花青素反射指数1(ARI_1)(1/B49)-(1/B93)[43]
      2花青素反射指数2(ARI_2)((1/B49)-(1/B93))*B121[43]
      3类胡萝卜素反射指数1(CRI_1)(1/B38)-(1/B49)[43]
      4类胡萝卜素反射指数2(CRI_2)(1/B38)-(1/B93)[43]
      5病害水分胁迫指数(DWSI)B49/B87[44]
      6增强型植被指数(EVI)2.5*(B121-B85)/(B121+6*B85-7.5*B26+1)[45]
      7绿色归一化植被指数(GNDVI)(B121-B49)/(B121+B49)[46]
      8归一化叶绿素指数(NDchl(B157-B95)/(B157+B95)[47]
      9归一化植被指数(NDVI)(B121-B87)/(B121+B87)[48]
      10光化学植被指数(PRI)(B42-B55)/(B42+B55)[49]
      11植被衰减指数(PSRI)(B87-B34)/B107[50]
      12植被水分指数(PWI)B149/B168[51]
      13红边拐点指数(REP)700+40*(((B85+B117)/2-B93)/(B105-B93))[52]
      14红边植被胁迫指数(RVSI)((B97+B107)/2)-B103[53]
      15简单比值指数(SR)B121/B87[54]
      16绿色植被指数(VIgreen)(B49-B87)/(B49+B87)[55]
      17Vogelmann红边指数1(VOG_1)B105/B99[56]
      18Vogelmann红边指数2(VOG_2)(B103-B107)/(B97+B101)[57]
      19修正归一化植被指数(MNDVI)(B109-B105)/(B109+B105)[57]
      20抗大气植被指数(ARVI)(B121-2*B84+B20)/(B121+2*B84-B20)[58]
      21土壤调节植被指数(SAVI)(1+0.5)*(B121-B84)/(B121+B84+0.5)[59]
      22优化型土壤调节植被指数(OSAVI)(1+0.16)*(B121-B84)/(B121+B84+0.16)[60]
      23改进型土壤调节植被指数(MSAVI)0.5*(B121+1-sqrt((2*B121+1)^2-8*(B121-B87)))[61]
      24绿色植被指数1(VIgreen_1)(B49-B84)/(B49+B84)[55]
      25简单比值指数1(SR_1)B99/B84[54]
      26简单比值指数2(SR_2)B141/B84[54]
      27简单比值指数3(SR_3)B164/B98[54]
      28植被水分指数1(PWI_1)B159/B166[51]
      29植被衰减指数1(PSRI_1)(B84-B34)/B99[50]
      30类胡萝卜素反射指数1_1(CRI1_1)(1/B34)-(1/B49)[44]
      31类胡萝卜素反射指数2_1(CRI2_1)(1/B40)-(1/B99)[44]
      32绿色归一化植被1(GNDVI_1)(B141-B49)/(B141+B49)[46]
    • Table 4. Classification results of 20 tree species under original spatial resolution

      View table
      View in Article

      Table 4. Classification results of 20 tree species under original spatial resolution

      分类方法变量类型变量数目特征变量整体精度Kappa
      随机森林原始波段9B34、B40、B49、B56、B84、B99、B159、B147、B1660.450.41
      植被指数12ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、RVSI、MNDVI、PWI_1、CRI1_1、CRI2_10.470.43

      原始波段+

      植被指数

      17B34、B40、B84、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、MNDVI、PRI、PSRI、PWI_1、REP0.500.46
      支持向量机原始波段9B34、B40、B49、B56、B84、B99、B159、B147、B1660.430.38
      植被指数12ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、RVSI、MNDVI、PWI_1、CRI1_1、CRI2_10.470.42

      原始波段+

      植被指数

      17B34、B40、B84、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、MNDVI、PRI、PSRI、PWI_1、REP0.500.46
    • Table 5. Impact of different spatial resolutions on the classification results of 10 tree species

      View table
      View in Article

      Table 5. Impact of different spatial resolutions on the classification results of 10 tree species

      分类方法空间分辨率变量数量特征变量整体精度Kappa
      支持向量机0.12 m9B34、B40、B166、ARI1、PRI、PSRI、MNDVI、PWI_1、DWSI4_20.590.52
      0.5 m10B34、B40、B166、ARI2、CRI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.70.65
      1 m10B34、B40、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.760.72
      1.5 m9B34、B40、ARI2、CRI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.770.73
      2 m8B34、B49、ARI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.760.72
      2.5 m10B34、B49、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.790.75
      3 m9B34、B49、B166、ARI2、PRI、PSRI、REP、PWI_1、CRI1_10.790.76
      3.5 m9B34、B49、B166、ARI2、PRI、PSRI、REP、PWI_1、CRI1_10.770.74
      4 m10B34、B49、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.750.72
      随机森林0.12 m9B34、B40、B166、ARI1、PRI、PSRI、MNDVI、PWI_1、DWSI4_20.580.52
      0.5 m10B34、B40、B166、ARI2、CRI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.670.61
      1 m10B34、B40、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.720.68
      1.5 m9B34、B40、ARI2、CRI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.730.69
      2 m8B34、B49、ARI2、PRI、PSRI、PWI_1、DWSI4_2、SR_30.730.69
      2.5 m10B34、B49、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.770.74
      3 m9B34、B49、B166、ARI2、PRI、PSRI、REP、PWI_1、CRI1_10.770.73
      3.5m9B34、B49、B166、ARI2、PRI、PSRI、REP、PWI_1、CRI1_10.770.72
      4 m10B34、B49、B166、ARI2、CRI2、PRI、PSRI、REP、PWI_1、DWSI4_20.740.70
    • Table 6. Impact of different spatial resolutions on the classification results of 15 tree species

      View table
      View in Article

      Table 6. Impact of different spatial resolutions on the classification results of 15 tree species

      分类

      方法

      空间

      分辨率

      变量数量特征变量整体精度Kappa

      支持

      向量机

      0.12 m14B34、B40、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、PWI_1、CRI1_1、CRI2_1、SR_3、GNDVI_10.560.51
      0.5 m15B34、B40、B84、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.670.63
      1 m16B34、B40、B84、B93、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.700.67
      1.5 m15B34、B40、B56、B84、B166、ARI1、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.730.71
      2 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.730.71
      2.5 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.730.71
      3 m13B34、B49、B99、B166、ARI1、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.740.72
      3.5 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.730.72
      4 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.710.69

      随机

      森林

      0.12 m14B34、B40、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、PWI_1、CRI1_1、CRI2_1、SR_3、GNDVI_10.550.50
      0.5 m15B34、B40、B84、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.630.59
      1 m16B34、B40、B84、B99、B166、ARI1、ARI2、CRI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.670.64
      1.5 m15B34、B40、B56、B84、B166、ARI1、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_1、GNDVI_10.690.66
      2 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.700.67
      2.5 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.710.68
      3 m13B34、B49、B99、B166、ARI1、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.730.70
      3.5 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.700.67
      4 m12B34、B49、B56、B166、ARI2、DWSI4、PRI、PSRI、REP、PWI_1、CRI1_1、CRI2_10.680.64
    • Table 7. Impact of different spatial resolutions on the classification results of 20 tree species

      View table
      View in Article

      Table 7. Impact of different spatial resolutions on the classification results of 20 tree species

      分类

      方法

      空间

      分辨率

      变量数量特征变量整体精度Kappa

      支持

      向量机

      0.12 m17B34、B40、B56、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、MNDVI、PRI、PSRI、PWI_1、REP0.500.46
      0.5 m16B34、B40、B84、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、PRI、PSRI、PWI_1、REP0.620.59
      1 m13B34、B40、B49、B166、ARI2、CRI1_1、CRI2_1、DWSI4、GNDVI_1、PRI、PSRI、PWI_1、REP0.660.64
      1.5 m12B34、B40、B99、B166、ARI2、CRI1_1、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.680.66
      2 m12B34、B40、B49、B166、ARI2、CRI1_1、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.700.67
      2.5 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.700.67
      3 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.710.69
      3.5 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.690.66
      4 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.670.64

      随机

      森林

      0.12 m17B34、B40、B56、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、MNDVI、PRI、PSRI、PWI_1、REP0.500.46
      0.5 m16B34、B40、B84、B99、B166、ARI1、ARI2、CRI1_1、CRI2、CRI2_1、DWSI4、GNDVI_1、PRI、PSRI、PWI_1、REP0.580.55
      1 m13B34、B40、B49、B166、ARI2、CRI1_1、CRI2_1、DWSI4、GNDVI_1、PRI、PSRI、PWI_1、REP0.610.58
      1.5 m12B34、B40、B99、B166、ARI2、CRI1_1、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.630.60
      2 m12B34、B40、B49、B166、ARI2、CRI1_1、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.660.63
      2.5 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.680.66
      3 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.700.68
      3.5 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.680.65
      4 m11B34、B40、B99、B166、ARI2、CRI2_1、DWSI4、PRI、PSRI、PWI_1、REP0.670.64
    • Table 8. Average classification results of 10, 15, and 20 tree species at different scales

      View table
      View in Article

      Table 8. Average classification results of 10, 15, and 20 tree species at different scales

      分类方法空间分辨率平均整体精度平均Kappa
      支持向量机0.12 m0.550.50
      0.5 m0.660.62
      1 m0.710.68
      1.5 m0.730.70
      2 m0.730.70
      2.5 m0.740.71
      3 m0.750.72
      3.5 m0.730.71
      4 m0.710.68
      随机森林0.12 m0.540.49
      0.5 m0.630.58
      1 m0.670.63
      1.5 m0.680.65
      2 m0.700.66
      2.5 m0.720.69
      3 m0.730.70
      3.5 m0.720.68
      4 m0.700.66
    Tools

    Get Citation

    Copy Citation Text

    Xiangshan ZHOU, Wunian YANG, Ke LUO, Hongyi PIAO, Tao ZHOU, Jie ZHOU, Xiaolu TANG. Research on UAV Hyperspectral of Tree Species Classification based on Machine Learning Algorithms and Spatial Resolution Adjustment[J]. Remote Sensing Technology and Application, 2024, 39(4): 880

    Download Citation

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

    Category:

    Received: Dec. 10, 2022

    Accepted: --

    Published Online: Jan. 6, 2025

    The Author Email:

    DOI:10.11873/j.issn.1004-0323.2024.4.0880

    Topics