林业科学 ›› 2021, Vol. 57 ›› Issue (5): 119-129.doi: 10.11707/j.1001-7488.20210511
由珈齐,李明泽*,范文义,全迎,王斌,莫祝坤,祝子枭
收稿日期:
2020-04-10
出版日期:
2021-05-25
发布日期:
2021-07-09
通讯作者:
李明泽
基金资助:
Jiaqi You,Mingze Li*,Wenyi Fan,Ying Quan,Bin Wang,Zhukun Mo,Zixiao Zhu
Received:
2020-04-10
Online:
2021-05-25
Published:
2021-07-09
Contact:
Mingze Li
摘要:
目的: 探讨随机森林、支持向量机分类器下机载高光谱影像和激光雷达点云数据源对林分类型识别的影响,并检验叶绿素在林分类型识别中的作用,为提高林分类型分类精度提供科学依据,为森林资源管理和监测提供技术支持。方法: 以东北林业大学帽儿山实验林场老山施业区为研究区,以机载高光谱影像和激光雷达点云为数据源,在多尺度影像分割基础上,从高光谱影像中提取光谱、纹理和叶绿素指数等特征,从LiDAR点云中提取高度、强度等特征。通过随机森林的特征选择,选取重要性较高的特征变量,在随机森林和支持向量机分类器下,以影像分割数据为试验样本,设置6种分类方案(随机森林分类器下高光谱影像与激光雷达点云数据结合、高光谱影像数据、激光雷达点云数据,支持向量机分类器下高光谱影像与激光雷达点云数据结合、高光谱影像数据、激光雷达点云数据),对阔叶混交林、樟子松林、落叶松林、红松林和蒙古栎林5种林分类型进行识别,比较不同分类器下不同数据源的分类效果。结果: 高光谱影像数据共提取34个特征变量,激光雷达点云数据共提取72个特征变量,经特征选择后,高光谱影像数据和激光雷达点云数据各选取11个重要性较高的特征(共22个),其中高光谱影像数据提取的归一化植被指数(NDVI)重要性最大。6种分类方案中,随机森林分类器下高光谱影像与激光雷达点云数据结合的分类精度最高(88.02%),支持向量机分类器下激光雷达点云数据的分类精度最低(76.19%)。多源数据协同的平均分类精度(86.22%)高于单源数据(79.98%),随机森林分类器的平均分类精度(82.92%)高于支持向量机分类器(81.19%)。叶绿素指数参与分类后,分类精度提高约3.32%。5种林分类型中,阔叶混交林分类效果最好,平均分类精度为92.62%,红松林分类效果最差,平均分类精度为49.67%。结论: 多数据源较单源数据可更好地提高分类精度,即2种数据协同可以提高林分类型识别精度;单一数据源相比,高光谱影像数据源的分类效果更好,光谱特征是林分类型识别的重要影响因子;林分类型识别时,不同机器学习模型相比,随机森林分类器较支持向量机分类器分类效果更优;叶绿素作为生物化学参数对林分类型识别有积极影响。
中图分类号:
由珈齐,李明泽,范文义,全迎,王斌,莫祝坤,祝子枭. 基于高光谱和激光雷达数据的林分类型识别[J]. 林业科学, 2021, 57(5): 119-129.
Jiaqi You,Mingze Li,Wenyi Fan,Ying Quan,Bin Wang,Zhukun Mo,Zixiao Zhu. Stand Type Identification Based on Hyperspectral and LiDAR Data[J]. Scientia Silvae Sinicae, 2021, 57(5): 119-129.
表1
高光谱数据提取特征计算公式"
特征类型 Feature type | 特征名称 Feature name | 描述或公式 Description or formula |
植被指数 Vegetation index | 归一化植被指数 Normalized differential vegetation index | 式中:ρNIR为近红外波段反射率,ρRED为红光波段反射率 In the formula: ρNIR is the reflectance of near-infrared band, ρRED is the reflectance of RED band |
增强型植被指数 Enhanced vegetation index | 式中:ρBLUE为蓝光波段反射率 In the formula: ρBLUE is the reflectance of BLUE band | |
比值植被指数 Ratio vegetation index | ||
大气阻抗植被指数 Atmospherically resistant vegetation index | ||
改进红边归一化植被指数 Modified red edge NDVI | 式中:ρ750为750波段反射率,ρ705为705波段反射率,ρ445为445波段反射率 In the formula: ρ750 is the reflectance of 750 band, ρ705 is the reflectance of 705 band, and ρ445 is the reflectance of 445 band | |
红边归一化植被指数 Red edge NDVI | ||
光化学反射指数 Photochemical reflectance index | 式中:ρ531为531波段反射率,ρ570为570波段反射率 In the formula: ρ531 is the reflectance of 531 band, ρ570 is the reflectance of 570 band | |
植物衰老反射指数 Plant senescence reflectance index | 式中:ρ680为680波段反射率,ρ500为500波段反射率 In the formula: ρ680 is the reflectance of 680 band, ρ500 is the reflectance of 500 band | |
Vogelmann红边指数1 Vogelmann red edge index 1 | 式中:ρ740为740波段反射率,ρ720为720波段反射率 In the formula: ρ740 is the reflectance of 740 band, ρ720 is the reflectance of 720 band | |
水波段指数 Water band index | 式中:ρ900为900波段反射率,ρ970为970波段反射率 In the formula: ρ900 is the reflectance of 900 band, ρ970 is the reflectance of 970 band | |
叶绿素指数 Chlorophyll index | MERIS陆地叶绿素指数 MERIS terrestrial chlorophyll index | 式中:ρ753.75为753.75波段反射率,ρ708.75为708.75波段反射率,ρ681.25为681.25波段反射率 In the formula: ρ753.75 is the reflectance of the 753.75 band, ρ708.75 is the reflectance of the 708.75 band, and ρ681.25 is the reflectance of the 681.25 band |
改进的MERIS陆地叶绿素指数 Modified MTCI | 式中:ρ750为750波段反射率,ρ710为710波段反射率,ρ680为680波段反射率 In the formula: ρ750 is the reflectance of the 750 band, ρ710 is the reflectance of the 710 band, and ρ680 is the reflectance of the 680 band |
表2
各林分样本数量"
项目 Item | 阔叶混交林 Broad-leaved mixed forest | 樟子松林 P. sylvestris var. mongolica forest | 落叶松林 L. gmelinii forest | 红松林 P. koraiensis forest | 蒙古栎林 Q. mongolica forest | 总计 Total |
训练样本Training sample | 1 358 | 374 | 224 | 74 | 47 | 2 077 |
检验样本Validation sample | 453 | 125 | 75 | 25 | 16 | 692 |
样本量Sample size | 1 810 | 499 | 299 | 98 | 63 | 2 769 |
图4
选取的特征重要性排序 a、b、c分别为激光雷达点云、高光谱影像、激光雷达点云+高光谱影像保留的特征变量。AII_90%为所有回波累计强度的90%;H1_max为首次回波高度的最大值;I1_1%为首次回波强度的1%;I_min为所有回波强度的最小值;I_max为所有回波强度的最大值;I1_cv为首次回波强度的偏度;I_90%为所有回波强度的90%;H_mean为所有回波高度的均值;AIH_99%为所有回波累计高度的99%;I_95%为所有回波强度的95%;H_95%为所有回波高度的95%;green_range为绿色波段卷积核范围内平均灰度范围;blue_mean为蓝色波段卷积核范围内平均灰度;green_mean为绿色波段卷积核范围内平均灰度;red_range为红色波段卷积核范围内平均灰度范围。"
表3
随机森林和支持向量机分类器的分类结果"
指标Indicator | 随机森林RF | 支持向量机SVM | ||||||
高光谱影像+激光雷达点云 Hyperspectral+ LiDAR | 高光谱影像 Hyperspectral | 激光雷达点 云LiDAR | 高光谱影像+激光雷达点云 Hyperspectral+ LiDAR | 高光谱影像 Hyperspectral | 激光雷达 点云LiDAR | |||
总体精度OA(%) | 88.02 | 84.13 | 76.62 | 84.42 | 82.97 | 76.19 | ||
Kappa系数Kappa coefficient | 0.77 | 0.69 | 0.52 | 0.71 | 0.68 | 0.53 | ||
阔叶混交林 Broad-leaved mixed forest | 生产者精度PA (%) | 95.32 | 93.60 | 85.89 | 94.13 | 93.78 | 85.68 | |
用户精度UA (%) | 94.07 | 92.78 | 92.51 | 92.05 | 93.16 | 91.17 | ||
落叶松林 L. gmelinii forest | 生产者精度PA (%) | 76.81 | 61.54 | 52.31 | 71.67 | 62.50 | 56.36 | |
用户精度UA (%) | 70.67 | 46.38 | 45.33 | 57.33 | 53.33 | 41.33 | ||
樟子松林 P. sylvestris var. mongolica forest | 生产者精度PA (%) | 73.29 | 66.04 | 50.88 | 65.82 | 66.67 | 55.37 | |
用户精度UA (%) | 85.60 | 80.77 | 46.03 | 83.20 | 75.20 | 53.60 | ||
红松林 P. koraiensis forest | 生产者精度PA (%) | 72.22 | 66.67 | 70.59 | 76.47 | 52.38 | 45.45 | |
用户精度UA (%) | 54.17 | 52.17 | 50.00 | 54.17 | 45.83 | 41.67 | ||
蒙古栎林 Q.mongolicaforest | 生产者精度PA (%) | 81.82 | 90.91 | 87.50 | 53.33 | 47.06 | 53.85 | |
用户精度UA (%) | 64.29 | 71.43 | 50.00 | 50.00 | 50.00 | 43.75 |
陈永健. 2014. 半监督支持向量机分类方法研究. 西安: 陕西师范大学硕士学位论文. | |
Chen Y J. 2014. Research on classification methods of semi-supervised support vector machines. Xi'an: MS thesis of Shaanxi Normal University. [in Chinese] | |
丁壮, 姚余君, 宋日清. 帽儿山实验林场老山地区森林资源的历史变迁. 东北林业大学学报, 2006, 34 (3): 87- 89.
doi: 10.3969/j.issn.1000-5382.2006.03.035 |
|
Ding Z , Yao Y J , Song R Q . Historical changes of forest resources in Laoshan area of Mao'ershan experimental forest farm. Journal of Northeast Forestry University, 2006, 34 (3): 87- 89.
doi: 10.3969/j.issn.1000-5382.2006.03.035 |
|
董恒, 孟庆野, 王金梁, 等. 一种改进的叶绿素提取植被指数. 红外与毫米波学报, 2012, 31 (4): 336- 341. | |
Dong H , Meng Q Y , Wang J L , et al. An improved chlorophyll extraction vegetation index. Journal of Infrared and Millimeter Wave, 2012, 31 (4): 336- 341. | |
樊雪. 2016. 机载PHI高光谱数据的森林类型精细识别方法研究. 北京: 中国林业科学研究院硕士学位论文. | |
Fan X. 2016. Research on fine recognition method of forest types based on airborne PHI hyperspectral data. Beijing: MS thesis of Chinese Academy of Forestry. [in Chinese] | |
皋厦, 申鑫, 代劲松, 等. 结合LiDAR单木分割和高光谱特征提取的城市森林树种分类. 遥感技术与应用, 2018, 33 (6): 1073- 1083. | |
Gao S , Shen X , Dai J S , et al. Classification of urban forest tree species combined with LiDAR single tree segmentation and hyperspectral feature extraction. Remote Sensing Technology and Application, 2018, 33 (6): 1073- 1083. | |
黄衍, 查伟雄. 随机森林与支持向量机分类性能比较. 软件, 2012, 33 (6): 107- 110.
doi: 10.3969/j.issn.1003-6970.2012.06.038 |
|
Huang Y , Zha W X . Comparison of classification performance between random forests and support vector machine. Software, 2012, 33 (6): 107- 110.
doi: 10.3969/j.issn.1003-6970.2012.06.038 |
|
李波. 2017. 机载激光点云和高光谱影像融合的城市地物分类研究. 武汉: 武汉大学硕士学位论文. | |
Li B. 2017. Classification of urban ground objects based on fusion of airborne laser point clouds and hyperspectral images. Wuhan: MS thesis of Wuhan University. [in Chinese] | |
李明泽, 付瑜, 于颖, 等. 基于多时相SAR数据和SPOT数据的盘古林场林分类型识别. 植物研究, 2016, 36 (4): 613- 619, 626. | |
Li M Z , Fu Y , Yu Y , et al. Stand type identification of Pangu forest farm based on multi-temporal SAR data and SPOT data. Plant Research, 2016, 36 (4): 613- 619, 626. | |
刘丽娟, 庞勇, 范文义, 等. 机载LiDAR和高光谱融合实现温带天然林树种识别. 遥感学报, 2013, 17 (3): 679- 695. | |
Liu L J , Pang Y , Fan W Y , et al. Airborne LiDAR and hyperspectral fusion for tree species identification in temperate natural forests. Journal of Remote Sensing, 2013, 17 (3): 679- 695. | |
刘舒, 姜琦刚, 马玥, 等. 基于多目标遗传随机森林特征选择的面向对象湿地分类. 农业机械学报, 2017, 48 (1): 119- 127. | |
Liu S , Jiang Q G , Ma Y , et al. Object-oriented wetland classification based on multi-objective genetic random forest feature selection. Journal of Agricultural Machinery, 2017, 48 (1): 119- 127. | |
刘思涵, 尚夏明, 马婷. 基于随机森林特征选择的森林类型分类. 北京测绘, 2019, 33 (12): 1518- 1522. | |
Liu S H , Shang X M , Ma T . Classification of forest types based on random forest feature selection. Beijing Mapping, 2019, 33 (12): 1518- 1522. | |
刘毅, 杜培军, 郑辉, 等. 基于随机森林的国产小卫星遥感影像分类研究. 测绘科学, 2012, 37 (4): 194- 196. | |
Liu Y , Du P J , Zheng H , et al. Classification of domestic small satellite remote sensing images based on random forests. Surveying and Mapping Science, 2012, 37 (4): 194- 196. | |
刘怡君, 庞勇, 廖声熙, 等. 机载LiDAR和高光谱融合实现普洱山区树种分类. 林业科学研究, 2016, 29 (3): 407- 412.
doi: 10.3969/j.issn.1001-1498.2016.03.015 |
|
Liu Y J , Pang Y , Liao S X , et al. Airborne LiDAR and hyperspectral fusion to realize tree species classification in Pu'er mountain area. Forest Research, 2016, 29 (3): 407- 412.
doi: 10.3969/j.issn.1001-1498.2016.03.015 |
|
鲁续坤. 2018. 机载LiDAR和高光谱数据的树种分类及三维显示. 成都: 电子科技大学硕士学位论文. | |
Lu X K. 2018. Tree species classification and 3D display of airborne LiDAR and hyperspectral data. Chengdu: MS thesis of University of Electronic Science and Technology. [in Chinese] | |
毛学刚, 魏晶昱. 基于多源遥感数据的面向对象林分类型识别. 应用生态学报, 2017, 28 (11): 3711- 3719. | |
Mao X G , Wei J Y . Object-oriented stand type identification based on multi-source remote sensing data. Journal of Applied Ecology, 2017, 28 (11): 3711- 3719. | |
王婷. 2017. 高光谱遥感图像数据降维和分类算法研究. 银川: 北方民族大学硕士学位论文. | |
Wang T. 2017. Research on dimension reduction and classification algorithms for hyperspectral remote sensing image data. Yinchuan: MS thesis of Northern University for Nationalities. [in Chinese] | |
温小乐, 钟奥, 胡秀娟. 基于随机森林特征选择的城市绿化乔木树种分类. 地球信息科学学报, 2018, 20 (12): 1777- 1786.
doi: 10.12082/dqxxkx.2018.180310 |
|
Wen X L , Zhong A , Hu X J . Classification of urban greening tree species based on random forest feature selection. Journal of Geoscience, 2018, 20 (12): 1777- 1786.
doi: 10.12082/dqxxkx.2018.180310 |
|
谢珠利. 2019. 基于多源高分辨率遥感数据的人工林树种分类研究. 杭州: 浙江农林大学硕士学位论文. | |
Xie Z L. 2019. Classification of plantation tree species based on multi-source and high-resolution remote sensing data. Hangzhou: MS thesis of Zhejiang Agriculture and Forestry University. [in Chinese] | |
熊艳, 高仁强, 徐战亚. 机载LiDAR点云数据降维与分类的随机森林方法. 测绘学报, 2018, 47 (4): 508- 518. | |
Xiong Y , Gao R Q , Xu Z Y . Random forest method for dimension reduction and classification of airborne LiDAR point cloud data. Acta Surveying and Mapping Sinica, 2018, 47 (4): 508- 518. | |
颜春燕, 牛铮, 王纪华, 等. 光谱指数用于叶绿素含量提取的评价及一种改进的农作物冠层叶绿素含量提取模型. 遥感学报, 2005, 9 (6): 742- 750. | |
Yan C Y , Niu Z , Wang J H , et al. Spectral index for evaluation of chlorophyll content extraction and an improved model for extraction of chlorophyll content in crop canopy. Journal of Remote Sensing, 2005, 9 (6): 742- 750. | |
杨珺雯, 张锦水, 朱秀芳, 等. 随机森林在高光谱遥感数据中降维与分类的应用. 北京师范大学学报: 自然科学版, 2015, 51 (S1): 82- 88. | |
Yang J W , Zhang J S , Zhu X F , et al. Application of random forests in dimension reduction and classification of hyperspectral remote sensing data. Journal of Beijing Normal University: Natural Science Edition, 2015, 51 (S1): 82- 88. | |
姚明煌. 2014. 随机森林及其在遥感图像分类中的应用. 厦门: 华侨大学硕士学位论文. | |
Yao M H. 2014. Random forest and its application in remote sensing image classification. Xiamen: MS thesis of Overseas Chinese University. [in Chinese] | |
钟舒怡, 李向新, 柏叶辉, 等. 基于多尺度分割与光谱特征提取建成区的方法. 软件导刊, 2018, 17 (9): 180- 184. | |
Zhong S Y , Li X X , Bai Y H , et al. A method for extracting built-up regions based on multiscale segmentation and spectral features. Software Guide, 2018, 17 (9): 180- 184. | |
朱凌红, 周澎, 王忠民, 等. 高光谱数据与叶绿素含量及植被指数的相关性研究进展. 内蒙古民族大学学报: 自然科学版, 2014, 29 (1): 41- 44.
doi: 10.3969/j.issn.1671-0185.2014.01.014 |
|
Zhu L H , Zhou P , Wang Z M , et al. Research progress on correlation of hyperspectral data with chlorophyll content and vegetation index. Journal of Inner Mongolia University for Nationalities: Natural Science Edition, 2014, 29 (1): 41- 44.
doi: 10.3969/j.issn.1671-0185.2014.01.014 |
|
Andersen H E , Mcgaughey R J , Reutebuch S E . Estimating forest canopy fuel parameters using LiDAR data. Remote Sensing of Environment, 2005, 94 (4): 441- 449.
doi: 10.1016/j.rse.2004.10.013 |
|
Asner G P , Martin R E . Canopy phylogenetic, chemical and spectral assembly in a lowland Amazonian forest. New Phytol, 2011, 189 (4): 999- 1012.
doi: 10.1111/j.1469-8137.2010.03549.x |
|
Clément D , Clément M , Arnaud L B , et al. Semantic segmentation of forest stands of pure species combining airborne LiDAR data and very high resolution multispectral imagery. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 126, 129- 145. | |
Cortes C , Vapnik V . Support-vector networks. Machine Learning, 1955, 20, 273- 297. | |
Dash J , Curran P J . The MERIS terrestrial chlorophyll index. International Journal of Remote Sensing, 2004, 25 (23): 5003- 5013. | |
Lefsky M A , Cohen W B , Parker G G , et al. LiDAR remote sensing for ecosystem studies. BioScience, 2002, 52 (1): 19- 30.
doi: 10.1641/0006-3568(2002)052[0019:LRSFES]2.0.CO;2 |
|
Mountrakis M J , Ogole C . Support vector machines in remote sensing: a review. ISPRS Journal of Photogrammetry and Remote Sensing, 2011, 66 (3): 247- 259.
doi: 10.1016/j.isprsjprs.2010.11.001 |
|
Shi Y F , Andrew K S , Tiejun W , et al. Tree species classification using plant functional traits from LiDAR and hyperspectral data. International Journal of Applied Earth Observation and Geoinformation, 2018, 73, 207- 219.
doi: 10.1016/j.jag.2018.06.018 |
|
Wulder M A , Hall R J , Coops N C , et al. High spatial resolution remotely sensed data for ecosystem characterization. BioScience, 2004, 54 (6): 511- 521.
doi: 10.1641/0006-3568(2004)054[0511:HSRRSD]2.0.CO;2 |
|
Zhao Y J , Zeng Y , Zheng Z J , et al. Forest species diversity mapping using airborne LiDAR and hyperspectral data in a subtropical forest in China. Remote Sensing of Environment, 2018, 213, 104- 114.
doi: 10.1016/j.rse.2018.05.014 |
[1] | 潘颖,丁鸣鸣,林杰,代侨,郭赓,崔琳琳. 基于PROSAIL模型和多角度遥感数据的森林叶面积指数反演[J]. 林业科学, 2021, 57(4): 90-106. |
[2] | 李文博,吕振刚,黄选瑞,张志东. 河北省北部华北落叶松人工林立地指数空间分布预测[J]. 林业科学, 2021, 57(3): 79-89. |
[3] | 曾伟生,孙乡楠,王六如,王威,蒲莹. 基于机载激光雷达数据的森林蓄积量模型研建[J]. 林业科学, 2021, 57(2): 31-38. |
[4] | 陈芳,程献宝,黄安民,王学顺. 基于人工蜂群算法优化SVM的NIR杉木弹性模量预测[J]. 林业科学, 2021, 57(1): 161-168. |
[5] | 林达坤,黄世国,张飞萍,梁光红,吴松青,胡霞,王荣. 基于改进差分进化算法的鳞翅目昆虫图像识别方法[J]. 林业科学, 2020, 56(3): 73-81. |
[6] | 赵霖,张晓丽,吴艳双,张斌. 面向机载高光谱数据的3D-CNN亚热带森林树种分类[J]. 林业科学, 2020, 56(11): 97-107. |
[7] | 靳川,查天山,贾昕,田赟,周文君,卫腾宙. 毛乌素沙地沙柳光系统Ⅱ光保护机制和能量分配动态及其影响因子[J]. 林业科学, 2020, 56(10): 34-44. |
[8] | 栗旭升,李虎,陈冬花,刘玉锋,刘赛赛,刘聪芳,胡国庆. 联合GF-5与GF-6卫星数据的多分类器组合亚热带树种识别[J]. 林业科学, 2020, 56(10): 93-104. |
[9] | 张雷, 孙鹏森, 刘世荣. 川西亚高山森林不同恢复阶段生长季蒸腾特征[J]. 林业科学, 2020, 56(1): 1-9. |
[10] | 吴项乾,曹林,申鑫,汪贵斌,曹福亮. 基于无人机激光雷达的银杏人工林有效叶面积指数估测[J]. 林业科学, 2020, 56(1): 74-86. |
[11] | 毛学刚, 竹亮, 刘怡彤, 姚瑶, 范文义. 高空间分辨率影像与SAR数据协同特征面向对象林分类型识别[J]. 林业科学, 2019, 55(9): 92-102. |
[12] | 侯晓静, 明金科, 秦荣水, 朱霁平. 基于随机森林模型的交界域火灾风险分析[J]. 林业科学, 2019, 55(8): 194-200. |
[13] | 夏永杰, 庞勇, 刘鲁霞, 陈博伟, 董斌, 黄庆丰. 高精度DEM支持下的多时期航片杉木人工林树高生长监测[J]. 林业科学, 2019, 55(4): 108-121. |
[14] | 刘俊祥, 于永畅, 郎蓬蓬, 孙振元. 旱柳枝条叶绿体光化学特征的径向异质性[J]. 林业科学, 2019, 55(3): 36-42. |
[15] | 王好运, 吴峰, 朱小坤, 谢维斌. 叶型对马尾松幼苗生长及叶绿素荧光特征的影响[J]. 林业科学, 2019, 55(3): 183-192. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||