Please wait a minute...
img

官方微信

遥感技术与应用  2021, Vol. 36 Issue (5): 1199-1208    DOI: 10.11873/j.issn.1004-0323.2021.5.1199
遥感应用     
使用高分遥感影像获取塔里木河胡杨高度信息
杨雪峰1,2()
1.新疆师范大学地理科学与旅游学院,新疆 乌鲁木齐 830054
2.新疆维吾尔自治区重点实验室/新疆干旱区湖泊环境与资源实验室,新疆 乌鲁木齐 830054
Estimation Height of Populus Euphratica in Tarim River Using VHR Satellite Images
Xuefeng Yang1,2()
1.College of Geographic Science and Tourism,Xinjiang Normal University,Urumqi 830054,China
2.Xinjiang Laboratory of Lake Environment and Resources in Arid Zone,Urumqi 830054,China
 全文: PDF(5060 KB)   HTML
摘要:

森林高度在森林生态状况、生物量水平研究中是一个重要参数,目前存在的多种获取树高的遥感技术,都不同程度存在一些问题。塔里木河下游胡杨林作为干旱区内陆河流域荒漠生态系统的核心构件和重要生态恢复对象,了解其高度信息有助于科学评估塔河下游受损生态系统的恢复程度。使用高分辨率遥感影像,利用面向对象影像分析技术,获取单木尺度的胡杨树冠,并提取对应的光谱、纹理和几何特征;在使用消费级无人机获取的树高数据支持下,分别使用Linear、MLP(Multilayer Perceptron, MLP)、PACE 和SVR(SVM Regression, SVR)方法建立树高回归模型获取塔里木河下游胡杨高度信息。结果表明:①基于光谱、纹理和几何特征建立的树高回归模型R2为0.668 7,RMSE为0.942 6 m,说明结合使用高分辨率卫星遥感和无人机遥感技术可以用于获取单木尺度的胡杨树高;②当使用所有特征时,MLP、PACE和SVR回归模型的相关系数均大于0.81,其中PACE回归模型精度最高;③在单木尺度上,光谱特征中包含有较多的树高信息,其次是纹理特征。

关键词: 胡杨树高高分影像面向对象影像分析回归模型    
Abstract:

Forest height is an important parameter in the study of forest ecological and biomass. At present, there are many remote sensing technologies that can obtain tree height, but all have some problems. Populus euphratica in the lower reaches of the Tarim River is the core component and important ecological restoration object of the desert ecosystem in the inland river basin of the arid area.Taking Populus euphratica in the lower reaches of Tarim River as an example, using high-resolution remote sensing image and object-oriented image analysis technology, the canopy of Populus euphratica at single wood scale is obtained, and the corresponding spectral, texture and geometric features are extracted; with support of the tree height data acquired by UAV, the tree height regression model is established by linear, MLP, PACE and SVR respectively. The results show that: (1) the regression model R2 based on the spectral, texture and geometric features is 0.668 7 at the highest, and RMSE is 0.942 6 m, which indicates that the combination of VHR satellite remote sensing and UAV can be used to obtain the height of Populus euphratica at single wood scale; (2) When all features are used, the correlation coefficients of MLP, PACE and SVR regression models are greater than 0.81, and PACE regression archieve the highest accuracy; (3) On the scale of single wood, the spectral features contain more tree height information, followed by texture features.

Key words: Populus Euphratica    Tree height    Very-High-Resolution remote sensing    Object based image analysis    Regression model
收稿日期: 2020-04-26 出版日期: 2021-12-08
ZTFLH:  TP79  
基金资助: 国家自然科学基金项目(41761075)
作者简介: 杨雪峰(1972-),男,新疆乌鲁木齐人,副教授,主要从事干旱区资源环境遥感技术应用研究。E?mail:744157426@qq.com
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  
杨雪峰

引用本文:

杨雪峰. 使用高分遥感影像获取塔里木河胡杨高度信息[J]. 遥感技术与应用, 2021, 36(5): 1199-1208.

Xuefeng Yang. Estimation Height of Populus Euphratica in Tarim River Using VHR Satellite Images. Remote Sensing Technology and Application, 2021, 36(5): 1199-1208.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2021.5.1199        http://www.rsta.ac.cn/CN/Y2021/V36/I5/1199

图1  研究区
数据类型WorldView2无人机影像
获取时间2018年7月2018年8月
波段数43
飞行高度770 km100 m
地面分辨率

全色:0.5 m

多光谱:2 m

0.05 m
观测天顶角61.5°45°
表1  影像参数
图2  技术路线(by WV2 and UAV)
图3  胡杨树冠获取过程
图4  CHM获取过程(a)DSM (b)DEM (c)CHM)
特征说明R特征说明R
Brightness亮度-0.176Max_Green绿光最大值-0.35
Max_diff最大差异度量0.63Mean_NdviNDVI均值0.602
Quantile _NdviNDVI中位数0.601Max_NdviNDVI最大值0.539
Quantile _Nir近红外中位数0.361Max_Blue蓝光最大-0.354
Quantile _Blue蓝光中位数-0.525Max_Nir近红外最大值0.298
Quantile _Green绿光中位数-0.514GLCM.ContrastGLCM反差-0.416
Quantile _Red红光中位数-0.549GLCM.CorrelationGLCM相关性0.307
Standard Deviation _Blue蓝光标准差-0.331GLCM.DissimilarityGLCM异质性-0.457
Standard Deviation _Red红光标准差-0.327GLCM.EntropyGLCM熵0.277
Standard Deviation _NdviNDVI标准差0.279GLCM.HomogeneityGLCM同质性0.449
Standard Deviation _Green绿光标准差-0.281GLCM.MeanGLCM均值-0.407
Standard Deviation _Nir近红外标准差0.194GLDV.Ang.2GLDV角二阶矩0.457
Mean_Blue蓝光均值-0.567GLDV.ContrastGLDV反差-0.416
Mean_Green绿光均值-0.555GLDV.EntropyGLDV熵-0.454
Mean_Red红光均值-0.575GLDV.MeanGLDV均值-0.457
Mean_Nir近红外均值0.348Area面积0.413
表2  特征列表
图5  特征与树高散点图(Y轴为树高)
指标名称计算方法说明

皮尔逊相关系数

R(Pearson Correlation Coefficient)

i=1m(pi-pˉ)(ai-aˉ)i=1m(pi-pˉ)2i=1m(ai-aˉ)2

a是实测值;

p是预测值;

m是样本数量

平均绝对误差

MAE(Mean Absolute Error)

1mi=1m|pi-ai|

均方根误差

RMSE(Root Mean Squared Error)

1mi=1m(pi-ai)2

相对平方根误差

RRSE(Root Relative Squared Error)

i=1m(pi-ai)2i=1m(aˉ-ai)2

相对绝对误差

RAE(Relative Absolute Error)

i=1m|pi-ai|i=1m|aˉ-ai|
表3  评价指标
指标Linear RegressionMLP RegressionPACE RegressionSVR
R0.411 30.426 90.411 30.411 4
MAE1.168 71.155 31.168 71.162 5
RMSE1.467 91.456 31.467 91.474 8
RRSE88.836 7 %87.818 8 %88.836 7 %88.368 5 %
RAE91.086 3 %90.367 6 %91.086 3 %91.512 6 %
表4  模型比较(几何特征)
指标Linear RegressionMLP RegressionPACE RegressionSVR
R0.585 40.595 70.584 70.584 4
MAE1.027 81.0171.028 41.026
RMSE1.305 71.293 61.306 61.311 8
RRSE78.129 9 %77.306 1 %78.174 2 %77.986 9 %
RAE81.019 9 %80.269 %81.075 8 %81.397 7 %
表5  模型比较(纹理特征)
指标Linear RegressionMLP RegressionPACE RegressionSVR
R0.671 80.779 30.779 20.799 6
MAE0.9510.801 10.801 90.759 2
RMSE1.1931.009 31.009 50.969 7
RRSE72.292 1 %60.897 1 %60.958 2 %57.710 9 %
RAE74.027 6 %62.627 8 %62.640 9 %60.058 3 %
表6  模型比较(光谱特征)
指标Linear RegressionMLP RegressionPace RegressionSVR
R0.716 10.813 90.814 70.8121
MAE0.891 40.732 40.7320.736 5
RMSE1.124 20.935 70.933 90.940 7
RRSE67.761 5%55.671 3%55.638 3%55.981 2%
RAE69.756 8%58.062 9 %57.950 8%58.370 0%
表7  模型比较(几何特征+纹理特征+光谱特征)
图6  模型预测值与测量值
1 Chopping M,Nolin A,Moisen G G,et al.Forest canopy height from the Multi angle Imaging SpectroRadiometer (MISR) assessed with high resolution discrete return LiDAR[J]. Remote Sensing of Environment, 2009, 113(10): 2172-2185.
2 Dong Lixin. New development of forest canopy height remote sensing[J]. Remote Sensing Technology and Application, 2016, 31(5): 833-845.
2 董立新.林分平均高度卫星遥感新进展[J].遥感技术与应用, 2016, 31(5): 833-845.
3 Levick S R, Rogers K H. LiDAR and object-based image analysis as tools for monitoring the structural diversity of savanna vegetation[J]. Int’l Archives for the Photo- grammetry, Remote Sensing and Spatial Information Sciences, 2006, 34(2): 532-538.
4 Lefsky M A. A global forest canopy height map from the moderate resolution imaging spectroradiometer and the geoscience laser altimeter system[J]. Geophysical Research Letters, 2010, 37(15): 450-468.
5 Omasa K, Urano Y, Oguma H, et al. Mapping of tree position of Larix leptolepis woods and estimation of Diameter at Breast Height (DBH) and biomass of the trees using range data measured by a portable scanning LiDAR[J]. Journal of the Remote Sensing Society of Japan, 2002, 22(5): 550-557.
6 Leeuwen M V, Nieuwenhuis M. Retrieval of forest structural parameters using LiDAR remote sensing[J]. European Journal of Forest Research, 2010, 129(4):749-770.
7 Bian Rui, Yanyun Nian , Gou Xiaohua, et al. Analysis of forest canopy height based on UAV LiDAR: a case study of Picea Crassifolia in the East and Central of the Qilian Mountains[J]. Remote Sensing Technology and Application, 2021, 36(3): 511-520.
7 边瑞,年雁云, 勾晓华, 等. 基于无人机激光雷达的森林冠层高度分析[J]. 遥感技术与应用, 2021, 36(3): 511-520.
8 Liu He, Gu Lingjia, Ren Ruizhi. Research progress of forest parameter acquisition based on UAV remote sensing technology[J]. Remote Sensing Technology and Application, 2021, 36(3): 489-501.
8 刘鹤, 顾玲嘉, 任瑞治. 基于无人机遥感技术的森林参数获取研究进展[J]. 遥感技术与应用, 2021, 36(3): 489-501.
9 Sarron J, Malézieux É, Sané C A B, et al. Mango yield mapping at the orchard scale based on tree structure and Land cover assessed by UAV[J]. Remote Sensing. 2018, 15(10):1900-1915. Doi: 10.3390/rs10121900
doi: 10.3390/rs10121900
10 Yang Kun, Zhao Yanling, Zhang Jianyong, et al. Tree height extraction using High-resolution imagery acquired from an Unmanned Aerial Vehicle(UAV). Journal of Beijing Forestry University[J]. 2017, 39(8): 17-23.杨坤, 赵艳玲, 张建勇, 等. 利用无人机高分辨率影像进行树木高度提取[J].北京林业大学学报, 2017, 39(8): 17-23.
11 Cloude S R. Robust Parameter Estimation using dual baseline polarimetric SAR Inter-ferometry[C]∥ Geoscience and Remote Sensing Symposium, 2002, 2, 838-840.
12 Cloude S R, Papathanassiou K P. Three-stage inversion process for polarimetric SAR interferometry[J]. IEE Proceedings-Radar, sonar and navigation, 2003, 150(3): 125-134.
13 Cloude S R. Polarization coherence tomography[J]. Radio Science, 2006, 41(4): 495-507.
14 Tanase M A, Panciera R, Lowell K, et al. Estimation of forest biomass from L-band polarimetric decomposition components[C]∥ 2013 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), IEEE, 2013: 949-952.
15 Tan Lulu, Chen Bing, Yang Ruliang. Improved three-stage algorithm of rree height retrieval with PolInSAR data[J]. Journal Of System Simulation, 2010, 22(4): 996-999.
15 谈璐璐, 陈兵, 杨汝良. 利用POLInSAR 数据反演植被高度的改进三阶段算法[J].系统仿真学报, 2010, 22(4): 996-999.
16 Park W, Sohn H G, Heo J. Estimation of forest canopy height using ortho image-refined digital elevation models[J]. Landscape and Ecological Engineering, 2015,11(1): 73-86.
17 Liu Xiaomeng, Chang Zhanqiang, Zhang Jingfa, et al. Investigation on decorrelation in processing of D-InSAR[J]. Journal Of Hebei Normal University (Natural Science Edition), 2007, 31(2): 260-263.
17 刘晓萌, 常占强, 张景发, 等. D-InSAR 处理中失相干问题的研究[J]. 河北师范大学学报(自然科学版), 2007, 31(2): 260-263.
18 Li Zhen, Guo Ming, Wang Zhongqiong, et al. Forest-height Inversion using repeat-pass spaceborne PolInSAR data[J]. Scientia Sinica(Terrae), 2014, (4): 680-692.李震, 郭明, 汪仲琼, 等. 星载重轨极化干涉SAR 反演森林植被高度[J]. 中国科学: 地球科学, 2014, (4): 680-692.
19 Han Xuefeng. A study on the dxtraction of the stand description factors from high resolution remote sensing images[D]. Fujian:FuJian Normal University, 2008.
19 韩学锋.基于高分辨率遥感林分调查因子的提取研究[D].福建:福建师范大学, 2008.
20 Zhou Zhiqiang, Yue Cairong, Xu Tianshu, et al. Review on forest height estimation using remote sensing[J]. Modern Agricultural Science and Technology,2012, 31(2): 198-199.
20 周志强, 岳彩荣, 徐天蜀, 等. 森林高度遥感估测研究综述[J].现代农业科技, 2012, 31(2):198-199.
21 Cui Shaowei,Fan Wenyi,Jin Sen,et al. Extraction of individual tree height using QuickBird images based on Tree Shadow[J]. Journal of Northeast Forestry University,2011,39(2): 47-50.
21 崔少伟, 范文义,金森,等.基于树影与快鸟图像的单木树高提取[J].东北林业大学学报, 2011, 39(2):47-50.
22 Ni Wenjian, Zhang Dafeng, Wang Yao, et al. Extraction of forest height by using GF-2 cross track stereo images[J]. Journal of Remote Sensing, 2017, 22(3): 24-31.
22 倪文俭, 张大凤, 汪垚, 等. 高分二号异轨立体数据的森林高度提取[J]. 遥感学报, 2017, 22(3): 24 -31.
23 Ling Chengxing, Ju Hongbo, Liu Hua, et al. A method of mean forest height fast extraction from GeoEye-1 stereo pairs data[J]. Geomatics& Spatial Information Technology, 2018, 41(3): 7-10.
23 凌成星, 鞠洪波, 刘华, 等. 一种基于GeoEye-1立体像对数据的快速提取平均树高方法[J].测绘与空间地理信息, 2018, 41(3): 7-10.
24 Peter T W, Philip A T, Brian R S. Estimation of forest structural parameters using 5 and 10 meter SPOT-5 satellite data[J]. Remote Sensing of Environment, 2009, 113(4), 2019–2036.
25 Donoghue D N M, Watt P J. Using LiDAR to compare forest height estimates from IKONOS and Landsat ETM+ data in Sitka Spruce plantation forests[J]. International Journal of Remote Sensing, 2006, 27(11): 2161-2175.
26 Haralick R M, Shanmugam K, Dinstein I. Textural features for image classification[J]. IEEE Transactions on Systems, Man and Cybernetics, 1973, 3(2): 610–621.
27 Chen G, Hay G J, Castilla G, et al. A multiscale geographic object-based image analysis to estimate LiDAR-measured forest canopy height using Quickbird imagery[J]. International Journal of Geographical Information Science, 2011, 25(6): 877-893.
28 Gomes M F, Maillard P. Using spectral and textural features from RapidEye images to estimate age and structural parameters of cerrado vegetation[J]. International Journal of Remote Sensing, 2015, 36(11-12): 3058-3076.
29 Shamsoddini A, Trinder J C, Turner R. Pine plantation structure mapping using WorldView2 multispectral image[J]. International Journal of Remote Sensing, 2013, 34(11): 3986-4007.
30 Niu Ting. Remote sensing quantitative evaluation of vegetation restoration in the lower reaches of Tarim River[D]. Xinjiang: Xinjiang Agricultural University, 2014.
30 牛婷. 塔里木河下游植被恢复遥感定量评估研究[D].新疆: 新疆农业大学, 2014.
31 Cai Fuyan, Halik Umut , Abibul Arkin , et al. Tree height distribution and density of populus euphratica forest along the Argan Section in the lower reaches of Tarim River[J]. Ecology and Environment, 2008, 17(3):1086-1090.
31 蔡富艳,玉米提·哈力克, 艾尔肯·艾白不拉, 等. 塔里木河下游阿拉干断面胡杨树高的分布和密度[J].生态环境, 2008, 17(3):1086-1090.
32 Wan Hongmei, Li Xia, Dong Daorui, et al. Abstraction and analysis of tree-crown of forest land based on QuickBird image in the lower reaches of Tarim River[J]. Acta Botanica Boreali-Occidentalia Sinica, 2011, 31(9): 1878 -1885.
32 万红梅,李霞,董道瑞,等.塔里木河下游林地树冠QuickBird 影像信息提取与分析[J]. 西北植物学报, 2011, 31(9): 1878-1885.
33 Jiang Hua. Landuse classification of high-resolution remote sensing image based on ecognition object-oriented technology[J]. Straits Science, 2014(8): 12-17.
33 江华.基于eCognition面向对象技术的高分辨率遥感影像土地利用分类—以福州琅岐岛为例[J].海峡科学, 2014(8): 12-17.
34 Song Yang, Li Chang Hui, Lin Hong. Application of object-oriented ecognition remote sensing image classification and recognition technology[J]. Geospatial Information,2012,10(2): 9,76-78.
34 宋杨,李长辉,林鸿.面向对象的eCognition遥感影像分类识别技术应用[J]. 地理空间信息,2012,10(2):9,76-78.
35 eCognition. eCognition Tutorial[EB/OL],, 2020-6-12.
36 Sadiq M. Prediction of software project effort using Linear Regression model[J]. International Journal of Information & Electronics Engineering, 2013, 3(3): 262-265.
37 Yong Wang, Ian H W. Modeling for optimal probability prediction[C]∥ Proceedings of the Nineteenth International Conference in Machine Learning,Sydney,Australia,2002, 650-657.
38 Rodrigues é O, Pinheiro V H A, Liatsis P, et al. Machine learning in the prediction of cardiac epicardial and mediastinal fat volumes[J]. Computers in Biology& Medicine, 2017: S0010482517300471.
39 Shevade S K, Keerthi S S, Bhattacharyya C, et al. Improvements to the SMO algorithm for SVM regression[C]∥ IEEE Transactions on Neural Networks, 1999.
40 Holmes G, Donkin A, Witten I H. WEKA: a Machine Learning Workbench[C]∥ Conference on Intelligent Information Systems. IEEE, 2002.
[1] 雷华锦,李弘毅,王建,郝晓华,赵宏宇,张娟. 基于环境信息和回归模型的青藏高原MODIS积雪面积比例产品制备[J]. 遥感技术与应用, 2020, 35(6): 1303-1311.
[2] 符海月,钱志友,张祎婷,王昭雅,洪娜娜. 基于多尺度的南京市城市用地分布空间决定因素[J]. 遥感技术与应用, 2020, 35(5): 1079-1088.
[3] 李越帅,郑宏伟,罗格平,杨辽,王伟胜,桂东伟. 集成U-Net方法的无人机影像胡杨树冠提取和计数[J]. 遥感技术与应用, 2019, 34(5): 939-949.
[4] 李琼琼, 柳云龙. 城市居民区土壤重金属含量高光谱反演研究[J]. 遥感技术与应用, 2019, 34(3): 540-546.
[5] 骆钰波, 黄洪宇, 唐丽玉, 陈崇成, 张浩. 基于地面激光雷达点云数据的森林树高、胸径自动提取与三维重建[J]. 遥感技术与应用, 2019, 34(2): 243-252.
[6] 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.
[7] 汪航,师茁. 基于MODIS时间序列数据的春尺蠖虫害遥感监测方法研究—以新疆巴楚胡杨为例[J]. 遥感技术与应用, 2018, 33(4): 686-695.
[8] 廖凯涛,齐述华,王成,王点. 结合GLAS和TM卫星数据的江西省森林高度和生物量制图[J]. 遥感技术与应用, 2018, 33(4): 713-720.
[9] 李兰,陈尔学,李增元,冯琦,赵磊. 合成孔径雷达森林树高和地上生物量估测研究进展[J]. 遥感技术与应用, 2016, 31(4): 625-633.
[10] 冯琦,陈尔学,李增元,李兰,赵磊. 机载XG波段双天线InSAR 数据森林树高估测方法[J]. 遥感技术与应用, 2016, 31(3): 551-557.
[11] 穆喜云,张秋良,刘清旺,庞勇,胡凯龙. 基于激光雷达的大兴安岭典型森林生物量制图技术研究[J]. 遥感技术与应用, 2015, 30(2): 220-225.
[12] 王金亮,程峰,王成,陈联君,王小花. 基于ICESat-GLAS数据估算复杂地形区域森林蓄积量潜力初探—以云南香格里拉县为例[J]. 遥感技术与应用, 2012, 27(1): 45-50.
[13] 毛德华,王宗明,罗玲,杨桄. 基于MODIS和AVHRR数据源的东北地区植被NDVI变化及其与气温和降水间的相关分析 [J]. 遥感技术与应用, 2012, 27(1): 77-85.
[14] 唐菲菲,阮志敏,刘星,张亚利. 基于机载激光雷达数据识别单株木的新方法[J]. 遥感技术与应用, 2011, 26(2): 196-201.
[15] 李 哲, 陈尔学, 王 建. 几种极化干涉SAR森林平均高反演算法的比较评价[J]. 遥感技术与应用, 2009, 24(5): 611-616.