Study on Rapid Extraction of Eucalyptus Vegetation Information in Guangxi Based on GEE

doi:10.13466/j.cnki.lyzygl.2019.05.010

Abstract

Abstract:

In order to further improve the efficiency of forest(plantation) vegetation extraction based on remote sensing image,this paper takes Landsat8 OIL as experimental data on the Google Earth Engine(GEE) platform,and uses supervised classification,support vector machine,maximum entropy model,random forest and decision tree classification based on the actual construction of the experimental area.Line extraction and comparison of various methods are made.On this basis,the area of Eucalyptus in Guangxi was extracted by decision tree,and the experimental results were validated by Unmanned aerial vehicle image and Google Earth Pro historical image.The experimental process and results show that remote sensing vegetation information can be extracted efficiently and quickly by using GEE.Among the five methods in this paper,the decision tree classification method achieves the best results.The overall accuracy and Kappa coefficients of Eucalyptus extraction in the experimental area are 0.82 and 0.85,respectively.At the same time,the area of Eucalyptus extracted by decision tree in Guangxi is in good agreement with the statistical data,which shows that the decision tree classification method constructed in this paper has a good consistency with the results of large area.Rapid extraction of vegetation cover information in complex mountainous areas is of reference significance.

Key words: Google Earth Engine, Guangxi, complex terrain, vegetation index, decision tree, remote sensing information extraction

CLC Number:

TP75
S792.39

LU Xianjian, HUANG Yuhui, YAN Hongbo, ZHOU Lv, WU Chenlong, ZHOU Bin, LUO Le. Study on Rapid Extraction of Eucalyptus Vegetation Information in Guangxi Based on GEE[J]. FOREST RESOURCES WANAGEMENT, 2019, (5): 52-60.

Figures/Tables 13

Fig.1

Fig.2

Fig.3

Tab.1

Tab.2

Information threshold table of vegetation index

指数类型	计算公式	所在层数
归一化植被指数(NDVI)	NDVI= $NIR - RED NIR + RED$	1
差值植被指数(DVI)	DVI=NIR-RED	2
垂直植被指数(PVI)	PVI=(0.939×NIR)-(0.344×RED)+0.09	3
土壤亮度植被指数(SAVI)	SAVI= $1.5 × (NIR - RED) NIR + RED + 0.5$	4
土壤调节植被指数(OSAVI)	OSAVI= $NIR - RED NIR + RED - 0.16$	4

Tab.2

Fig.4

Tab.3

Fig.5

Fig.6

Tab.4

Fig.7

Fig.8

Fig.9

References 24

[1]	黄国勤, 赵其国. 广西桉树种植的历史、现状、生态问题及应对策略[J]. 生态学报, 2014,34(18):5142-5152. doi: 10.5846/stxb201405100943
[2]	Sar1y1lmaz F B, Musaoğlu N, Uluğtekin N. Producing information for corine database by using classification method:a case study of sazlidere basin,istanbul[J]. ISPRS Annals of Photogrammetry,Remote Sensing & Spatial Information Sciences, 2017,4:357-362.
[3]	Chianga S H, Valdeza M, Chena C F. Forest tree species distribution mapping using Landsat satellite imagery and topographic variables with the maximum entropy method in Mongolia[J]. International Archives of the Photogrammetry,Remote Sensing and Spatial Information Sciences, 2016,8:593-596.
[4]	Zheng Baojuan, Myint S W, Thenkabail P S, et al. A support vector machine to identify irrigated crop types using time-series Landsat NDVI data[J]. International Journal of Applied Earth Observations & Geoinformation, 2015,34(1):103-112.
[5]	Karlson M, Ostwald M, Reese H, et al. Mapping tree canopy cover and aboveground biomass in sudano-sahelian woodlands using landsat8 and random forest[J]. Remote Sensing, 2015,7(8):10017-10041. doi: 10.3390/rs70810017
[6]	田静, 邢艳秋, 姚松涛, 等. 基于元胞自动机和BP神经网络算法的Landsat-TM 遥感影像森林类型分类比较[J]. 林业科学, 2017,53(2):26-34.
[7]	翟天林, 金贵, 邓祥征, 等. 植被信息的Landsat8卫星影像提取方法[J]. 测绘科学, 2016,41(10):126-131.
[8]	陈丽萍, 孙玉军. 基于不同决策树的面向对象林区遥感影像分类比较[J]. 应用生态学报, 2018,29(12):107-115.
[9]	徐晗泽宇, 刘冲, 王军邦, 等. Google Earth Engine平台支持下的赣南柑橘果园遥感提取研究[J]. 地球信息科学学报, 2018,20(3):396-404. doi: 10.12082/dqxxkx.2018.170553
[10]	乔海浪. 基于NDVI时间序列重构的经济型人工林时空分布信息提取研究[D]. 北京:中国科学院大学(中国科学院遥感与数字地球研究所), 2017.
[11]	袁琳. 张雷广西森林覆盖率达到62.31%[EB/OL]. 广西新闻网-广西日报, (2018-01-18)[2019-07-04]. https://v.gxnews.com.cn/a/16853738.
[12]	黄绍良. 广西桉树种植历史、现状、生态问题及对策分析[J]. 乡村科技, 2017(30):49-50.
[13]	郝斌飞, 韩旭军, 马明国, 等. Google Earth Engine在地球科学与环境科学中的应用研究进展[J]. 遥感技术与应用, 2018,33(4):600-611.
[14]	平亮, 熊高明, 谢宗强. 三峡库区桉树人工林封育后的群落特征和演替趋势[J]. 自然资源学报, 2009,24(9):1604-1615. doi: 10.11849/zrzyxb.2009.09.010
[15]	陶欢, 李存军, 周静平, 等. 基于高分1号影像的森林植被信息提取[J]. 自然资源学报, 2018. 33(6):1068-1079.
[16]	江曲图, 何俊昱, 王占山, 等. 基于LUR/BME的海岸带地区PM2.5时空特性研究[J]. 中国环境科学, 2017,37(2):424-431.
[17]	张静, 张翔, 田龙, 等. 西北旱区遥感影像分类的支持向量机法[J]. 测绘科学, 2017,42(1):49-52,58.
[18]	孙雪莲, 舒清态, 欧光龙, 等. 基于随机森林回归模型的思茅松人工林生物量遥感估测[J]. 林业资源管理, 2015(1):71-76.
[19]	郝泷, 陈永富, 刘华, 等. 基于纹理信息CART决策树的林芝县森林植被面向对象分类[J]. 遥感技术与应用, 2017,32(2):386-394.
[20]	Kundu A, Denis D, Patel N, et al. A Geo-spatial study for analyzing temporal responses of NDVI to rainfall[J]. Singapore Journal of Tropical Geography, 2017,39(1):107-116. doi: 10.1111/sjtg.12217
[21]	蒙诗栎, 庞勇, 张钟军, 等. WorldView-2纹理的森林地上生物量反演[J]. 遥感学报, 2017(5):812-824. doi: 10.11834/jrs.20176083
[22]	王小标, 谢顺平, 都金康. 水体指数构建及其在复杂环境下有效性研究[J]. 遥感学报, 2018,22(2):360-372.
[23]	潘仙昌. 南中亚热带过渡区较高海拔山地桉树无性系中试与评价[J]. 安徽农学通报, 2010,16(13):177-179.
[24]	罗来凤. 广西桉树人工林经济发展状况研究[J]. 绿色科技, 2016(13):57-58.

代码	类别	训练样本点/个
0	桉树	236
1	居民地	42
2	水系	36
3	耕地	54
4	其他植被	141
	合计	509

分类方法	分类精度	Kappa系数	提取面积/km²
监督分类	0.76	0.77	15.79
最大熵模型分类	0.74	0.71	18.44
支持向量机分类	0.79	0.78	13.90
随机森林分类	0.81	0.81	13.51
决策树分类	0.82	0.85	12.39

影像类型	地形类型	地形描述	吻合度/%
无人机影像(分辨率为0.2m)	滩涂	位于河流岸边滩涂,有农作物、灌木与成片桉树林等,可区分度高	84.2
无人机影像(分辨率为0.2m)	平地	位于植被覆盖率较高的平地,植被类型在3种以上,复杂度适中	81.7
Google Earth Pro 历史影像 (分辨率为0.5m)	山地	位于九龙森林公园,具有一定坡度的山地,桉树与其他类型植被交接,甚至混杂,复杂度较大	73.6
Google Earth Pro 历史影像 (分辨率为0.5m)	喀斯特地貌地区	位于广西东北部树种丰富的喀斯特地区,植被分布较杂乱,复杂度困难	72.7