Please wait a minute...
img

官方微信

遥感技术与应用  2020, Vol. 35 Issue (4): 855-863    DOI: 10.11873/j.issn.1004-0323.2020.4.0855
甘肃遥感学会专栏     
基于多源遥感影像的青海云杉和祁连圆柏分类
李萌(),年雁云(),边瑞,白艳萍,马金辉
兰州大学资源环境学院,甘肃 兰州 730000
Classification of Picea crassifolia and Sabina przewalskii based on Multi-source Remote Sensing Images
Meng Li(),Yanyun Nian(),Rui Bian,Yanping Bai,Jinhui Ma
College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China
 全文: PDF(3975 KB)   HTML
摘要:

青海云杉和祁连圆柏是祁连山自然保护区的优势种,提取两种类型树木的空间分布对保护区森林资源的管理和监测方面有重要意义。使用Sentinel-2A(S2)、Sentinel-1A(S1)、Landsat-8(L8)3种遥感影像及来自SRTM DEM的地形数据,基于随机森林分类方法,设置8种组合方案共22个特征变量,以祁连山东段的甘肃连城自然保护区为例,对青海云杉和祁连圆柏进行分类试验。结果表明:融合Sentinel-1A(S1)数据的VV和VH两种后向散射信息的精度最高,达到92.85%,比使用单一影像Landsat-8提高了11.64%。实验表明:结合多源遥感影像的不同波段信息是提高森林类型分类精度的有效手段,有助于复杂山区森林资源调查、植被信息提取等需求。

关键词: Sentinel-2A特征变量随机森林信息提取连城自然保护区    
Abstract:

Picea crassifolia and Sabina przewalskii are the dominant species in Liancheng Nature Reserve. Extracting the spatial distribution of two types of trees is of great significance for the management and monitoring of forest resources in the reserve. Based on the method of random forest,22 feature variables in eight combinations from Sentinel-2A (S2),Sentinel-1A (S1),Landsat-8 (L8) three remote sensing images and digital elevation model of SRTM DEM to classify Picea crassifolia and Sabina przewalskii in Liancheng Nature Reserve of Gansu Province.The results demonstrated that the accuracy of integrating VV and VH backscattering information of sentinels-1A (S1) was the highest,reaching 92.85%,which is 11.64% higher than that of single image Landsat-8. Experiments showed that combining different bands of multi-source remote sensing images is an effective means to improve the classification accuracy of forest types,which is beneficial to forest resource survey and vegetation information extraction in complex mountainous areas.

Key words: Sentinel-2A    Feature variables    Random forest    Information extraction    Liancheng National Nature Reserve
收稿日期: 2019-08-29 出版日期: 2020-09-15
ZTFLH:  S757  
基金资助: 中国科学院A类战略性先导科技专项(XDA20100102)
通讯作者: 年雁云     E-mail: limeng17@lzu.edu.cn;yynian@lzu.edu.cn
作者简介: 李萌(1994-),女,甘肃庆阳人,硕士研究生,主要从事遥感与GIS应用研究。E?mail:limeng17@lzu.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  
李萌
年雁云
边瑞
白艳萍
马金辉

引用本文:

李萌,年雁云,边瑞,白艳萍,马金辉. 基于多源遥感影像的青海云杉和祁连圆柏分类[J]. 遥感技术与应用, 2020, 35(4): 855-863.

Meng Li,Yanyun Nian,Rui Bian,Yanping Bai,Jinhui Ma. Classification of Picea crassifolia and Sabina przewalskii based on Multi-source Remote Sensing Images. Remote Sensing Technology and Application, 2020, 35(4): 855-863.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2020.4.0855        http://www.rsta.ac.cn/CN/Y2020/V35/I4/855

图1  研究区位置及样本点分布示意
数据重访周期/d分辨率/m波段数或波段名

Sentinel-2A

Landsat-8

Sentinel-1A

10

16

12

10/20/60

30/15

5×20

13

11

C 波段

表1  3种遥感数据的基本信息
分辨率/m波段名称中心波长/nm编号
10Blue490b2
Green560b3
Red665b4
NIR842b8
20Vegetation Red Edge1705b5
Vegetation Red Edge2740b6
Vegetation Red Edge3783b7
Narrow NIR865b8a
SWIR 11610b11
SWIR 22190b12
60Coastal aerosol443b1
Water vapour940b9
SWIR_cirrus1375b10
表2  Sentinel-2A波段信息
图2  优势树种不同图像对比
优势树种解译特征描述
青海云杉

坡向:主要分布在阴坡、半阴坡和河谷地带;

形状:顶部比较尖,树形又高又尖;

分布:呈大片密集状态紧凑分布;

色调:偏暗偏深

祁连圆柏

坡向:主要分布在阳坡;

形状:树高较低、顶部圆润,类似棒槌形状;

分布:呈零星状分布;

色调:较亮

表3  优势树种解译特征对比
图3  Landsat-8和Sentinel-2A影像中青海云杉和祁连圆柏的平均光谱特征
图4  技术流程
数据特征类型特征名称描述

Sentinel-2A

(10 bands)

光谱特征S2mean波段2-8、8a、11-12的均值
S2rviNIR/R
S2evi2.5×((NIR-R)/(NIR+6×R-7.5×B+1))
S2dviNIR-R
S2ndvi(NIR-R)/(NIR+R)
S2b78a(NIR2-RE3)/(NIR2+RE3)
S2b67(RE3-RE2)/(RE3+RE2)
S2b58a(NIR2-RE1)/(NIR2+RE1)
S2b56(RE2-RE1)/(RE2+RE1)
S2b57(RE2-RE1)/(RE2+RE1)
S2b68a(NIR2-RE2)/(NIR2+RE2)
S2b48a(NIR2-R)/(NIR2+R)

Sentinel-2A

(4 bands)

纹理特征PCA1

B、G、R、NIR基于PCA分析的

第一主成分

DEM地形特征ElevationDEM的高程值
Aspect基于DEM提取的坡向
Slope基于DEM提取的坡度
表4  用于第二层分类的特征变量
数据特征名称描述

Sentinel-2A

(10 bands)

S2mean波段2-8、8a、11-12的均值
S2rviNIR/R
S2evi2.5×((NIR-R)/(NIR+6×R-7.5×B+1))
S2dviNIR-R
S2ndvi(NIR-R)/(NIR+R)
S2b78a(NIR2-RE3)/(NIR2+RE3)
S2b67(RE3-RE2)/(RE3+RE2)
S2b58a(NIR2-RE1)/(NIR2+RE1)
S2b56(RE2-RE1)/(RE2+RE1)
S2b57(RE2-RE1)/(RE2+RE1)
S2b68a(NIR2-RE2)/(NIR2+RE2)
S2b48a(NIR2-R)/(NIR2+R)

Landsat-8

(7 bands)

L8mean波段1-7的均值
L8ndvi(NIR-R)/(NIR+R)
L8rviNIR/R
L8evi2.5×((NIR-R)/(NIR+6×R-7.5×B+1))
L8dviNIR-R
DEMElevationDEM的高程值
Aspect基于DEM提取的坡向
Slope基于DEM提取的坡度
Sentinel-1

VV

VH

VV极化方式的后向散射系数

VH极化方式的后向散射系数

表5  用于第三层分类的特征变量
图5  研究区林地提取结果及细节对比
组合数据总体精度/%Kappa系数
1S296.910.9312
2S2+DEM98.430.9681
3S2+L897.480.9486
4S2+S197.470.9487
表6  4种组合的林地提取精度
组合数据或特征变量总体精度/%Kappa系数
1L881.210.7240
2S288.040.8241
3S2+Elevation+Aspect+Slope90.540.8608
4S2+L889.590.8485
5S2+L8+Elevation+Aspect+Slope92.830.8957
6S2+L8+Elevation+Aspect+Slope+VV91.240.8724
7S2+L8+Elevation+Aspect+Slope+VH91.230.8722
8S2+L8+Elevation+Aspect+Slope+VV+VH92.850.8958
表7  8种组合的优势种分类精度
类别用户精度/%生产精度/%
青海云杉86.0389.79
祁连圆柏86.6190.12
表8  组合8的青海云杉和祁连圆柏的分类精度
图6  研究区优势树种分类结果及细节对比
图7  特征变量重要性排序
1 Wu Y S,Zhang X L. Object-based Tree Species Classification Using Airborne Hyperspectral Images and LiDAR Data[J].Forests,2020,11,32. doi:10.3390/f1101003.
doi: 10.3390/f1101003
2 Krzystek P,Serebryanyk A,Schnörr C,et al. Large-scale Ma-pping of Tree Species and Dead Trees in Šumava National Parkand Bavarian Forest National Park Using LiDAR and Multispectral Imagery[J].Remote Sensing,2020,12,661. doi:10.3390/rs12040661.
doi: 10.3390/rs12040661
3 Nezami S,Khoramshahi E,Nevalainen O,et al. Tree Species Classification of Drone Hyperspectral and RGB Imagery with Deep Learning Convolutional Neural Networks[J].Remote Sensing,2020,12,1070. doi:10.3390/rs12071070.
doi: 10.3390/rs12071070
4 Ren Chong,Ju Hongbo,Zhang Huaiqing,et al.Multi-source Data for Forest Land Type Precise Classification[J].Scientia SilvaeSinicae,2016,52(6):54-65.
4 任冲,鞠洪波,张怀清,等.多源数据林地类型的精细分类方法[J].林业科学,2016,52(6):54-65.
5 Li Mengying,Xing Yanqiu,Liu Meishuang,et al. Identifcation of Forest Type with Landsat-8 Image based on SVM[J]. Journal of Central South University of Forestry & Technology,2017,37(4):52-58.
5 李梦颖,邢艳秋,刘美爽,等.基于支持向量机的Landsat-8影像森林类型识别研究[J].中南林业科技大学学报,2017,37(4):52-58.
6 Tian Tian,Fan Wenyi,Lu Wei,et al. An Object-based Information Extraction Technology for Dominant Tree Species Group Types[J].Chinese Journal of Applied Ecology,2015(6):1665-1672.
6 田甜,范文义,卢伟,等.面向对象的优势树种类型信息提取技术[J].应用生态学报,2015(6):1665-1672.
7 Markus I,Francesco V,Clement A. First Experience with Sentinel-2 Data for Crop and Tree Species Classifications in CentralEurope[J]. Remote Sensing,2016,8(3):166. doi:10.3390/rs8030166.
doi: 10.3390/rs8030166
8 Zhu X L,Liu D S. Accurate Mapping of Forest Types Using Dense Seasonal Landsat Time-series[J]. ISPRS Journal of Photogra mmetry and Remote Sensing,2014,96:1-11.
9 Xie Z L,Chen Y L,Lu D S,et al. Classification of Land C-over,Forest,and Tree Species Classes with ZiYuan-3 Multispectral and Stereo Data[J].Remote Sensing,2019,11(2):164. doi:10.3390/rs11020164.
doi: 10.3390/rs11020164
10 Watanachaturaporn P,Arora M K,Varshney P K. Multisource Classification Using Support Vector Machines[J]. Photogrammetric Engineering & Remote Sensing,2015,74(2):239-246.
11 Forster M,Kleinschmit B. Significance Analysis of Different Types of Ancillary Geodata Utilized in a Multisource ClassificationProcess for Forest Identification in Germany[J]. IEEE Transactions on Geoscience & Remote Sensing,2014,52(5):3453-3463.
12 Xu X D,Li W,Ran Q,et al. Multisource Remote Sensing D-ata Classification based on Convolutional Neural Network[J]. IEEE Transactions on Geoscience & Remote Sensing,2017,(99):1-13.
13 Li J Y,Zhang H Y,Guo M,et al.Urban Classification by the Fusion of Thermal Infrared Hyperspectral and Visible Data [J]. Photogrammetric Engineering & Remote Sensing,2015,81(12):901-911.
14 Lei Guangbin,Li Ainong,Tan Jianbo,et al.Forest Types Mapp-ing in Mountainous Area Using Multi-source and Multi-temporalSatellite Images and DecisionTree Models[J].Remote Sensing Technology and Application,2016,31(1):31-41.
14 雷光斌,李爱农,谭剑波,等.基于多源多时相遥感影像的山地森林分类决策树模型研究[J].遥感技术与应用,2016,31(1):31-41.
15 Liu Gang,Zhang Yu,Zang Zhuo,et al.Extraction and Compar-ative Analysis of Forest Information of Multi-source Remote Sensing Data[J].Journal of Central South University of Forestry & Technology,2012,32(10):158-161.
15 刘刚,张雨,臧卓,等.多源遥感数据森林信息的提取和比较分析[J].中南林业科技大学学报,2012,32(10):158-161.
16 Wen Xiaole,Zhong Ao,Hu Xiujuan. The Classification of Urban Greening Tree Species based on Feature Selection of Random Forest[J]. Journal of Geo-information Science,2018,20(12):1777-1786.
16 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[J].地球信息科学学报,2018,20(12):1777-1786.
17 Jie Lü,Hao Ningyan,Li Chonggui,et al. Identification of For-est Type based on Random Froest and Texture Characteristic[J].Remote Sensing Information,2017,32(6):109-114.
17 吕杰,郝宁燕,李崇贵,等.利用随机森林和纹理特征的森林类型识别[J].遥感信息,2017,32(6):109-114.
18 Persson M,Lindberg E,Reese H. Tree Species Classification with Multi-temporal Sentinel-2 Data [J]. Remote Sensing,2018,10,1794. doi:10.3390/rs10111794.
doi: 10.3390/rs10111794
19 Li Ruonan,Guanglong Ou ,Dai Qinling,et al. Forest Types Classification in Shangri La,Northwest Yunnan with Landsat Time series Data based on Google Earth Engine[J]Journal of Southwest Forestry University(Natural Sciences),2020,40(5):1-12.
19 李若楠,欧光龙,代沁伶,等.基于GEE和Landsat时间序列数据的香格里拉森林类型分类研究[J].西南林业大学学报(自然科学版),2020,40(5):1-12.
20 Liu Sihan,Shang Xiaming,Ma Ting. The Classification of Forest Types. based on Feature Selection of Random Forest[J]. Beijing Surveying and Mapping,2019,33(12):1518-1522.
20 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[J].北京测绘,2019,33(12):1518-1522.
21 Bie Qiang,He Lei,Zhao Chuanyan,et al. Study on VegetationInformation Extraction based on Object-oriented Image Analysis[J]. Remote Sensing Technology and Application,2014,29(1):164-171.
21 别强,何磊,赵传燕.基于影像融合和面向对象技术的植被信息提取研究[J].遥感技术与应用,2014,29(1):164-171.
22 Wang Xiaoxue,Shen Huitao,Lin Tianmiao,et al.Improving the Forest Type Mapping Accuracy in Semiarid Mountainous Areas based on TM Image-Take the West Mountain of Beijing as an Example[J]. Journal of Natural Resources,2017,32(7):1217-1228.
22 王晓学,沈会涛,林田苗,等.利用多信息源提高半干旱地区TM影像的森林类型制图精度:以北京西部山区为例[J].自然资源学报,2017,32(7):1217-1228.
23 Liu Y A,Gong W S,Hu X Y,et al. Forest Type Identificati-on with Random Forest Using Sentinel-1A,Sentinel-2A,Multi-Temporal Landsat-8 and DEM Data [J]. Remote Sensing,2018,10(6):946. doi:10.3390/rs10060946.
doi: 10.3390/rs10060946
[1] 李宏达,高小红,汤敏. 基于CNN的不同空间分辨率影像土地覆被分类研究[J]. 遥感技术与应用, 2020, 35(4): 749-758.
[2] 张坤,刘乃文,高帅,赵书慧. 数据驱动的植被总初级生产力估算方法研究[J]. 遥感技术与应用, 2020, 35(4): 943-949.
[3] 李净,温松楠. 基于3种机器学习法的太阳辐射模拟研究[J]. 遥感技术与应用, 2020, 35(3): 615-622.
[4] 董超,赵庚星. 时序数据集构建质量对土地覆盖分类精度的影响研究[J]. 遥感技术与应用, 2020, 35(3): 558-566.
[5] 柴旭荣,李明,周义,王金风,田庆春. 影像的土地覆被快速分类[J]. 遥感技术与应用, 2020, 35(2): 315-325.
[6] 唐廷元,付波霖,何素云,娄佩卿,闭璐. 基于GF-1和Sentinel-1A的漓江流域典型地物信息提取[J]. 遥感技术与应用, 2020, 35(2): 448-457.
[7] 刘培,余志远,马威,韩瑞梅,陈正超,王涵,杨磊库. 基于地形信息的Landsat与Radarsat-2遥感数据协同分类研究[J]. 遥感技术与应用, 2019, 34(6): 1269-1275.
[8] 李哲,张沁雨,彭道黎. 基于高分二号遥感影像的树种分类方法[J]. 遥感技术与应用, 2019, 34(5): 970-982.
[9] 刘立, 刘勇. 基于优化分割分类层次体系的土地覆被分类制图方法探讨[J]. 遥感技术与应用, 2019, 34(1): 79-89.
[10] 谷晓天, 高小红, 马慧娟, 史飞飞, 刘雪梅, 曹晓敏. 复杂地形区土地利用/土地覆被分类机器学习方法比较研究[J]. 遥感技术与应用, 2019, 34(1): 57-67.
[11] 皋厦, 申鑫, 代劲松, 曹林. 结合LiDAR单木分割和高光谱特征提取的城市森林树种分类[J]. 遥感技术与应用, 2018, 33(6): 1073-1083.
[12] 高书鹏, 史正涛, 刘晓龙, 柏延臣. 基于高时空分辨率可见光遥感数据的热带山地橡胶林识别[J]. 遥感技术与应用, 2018, 33(6): 1122-1131.
[13] 韩涛, 潘剑君, 张培育, 曹罗丹. Sentinel-2A与Landsat-8影像在油菜识别中的差异性研究[J]. 遥感技术与应用, 2018, 33(5): 890-899.
[14] 陈伟民,张凌,宋冬梅,王斌,丁亚雄,许明明,崔建勇. 基于AdaBoost改进随机森林的高光谱图像地物分类方法研究[J]. 遥感技术与应用, 2018, 33(4): 612-620.
[15] 刘振波,邹娴,葛云健,陈健,曹雨濛. 基于高分一号WFV影像的随机森林算法反演水稻LAI[J]. 遥感技术与应用, 2018, 33(3): 458-464.