IMHE OpenIR
基于无人机倾斜摄影的亚高山针叶林地上生物量估测方法研究
Alternative TitleMethod of Above-ground Biomass Estimation of Subalpine Coniferous Forest Based on Unmanned Aerial Vehicle Oblique Photography
王枚梅
Subtype硕士
Thesis Advisor林家元
2018
Degree Grantor中国科学院大学
Place of Conferral北京
Degree Discipline地图学与地理信息系统
Keyword无人机倾斜摄影 地上生物量 点云 亚高山针叶林 树高
Abstract森林生物量指植物在单位面积上所积累的物质数量,其实质是绿色植物通过同化器官进行光合作用积累的有机质和能量。川西亚高山针叶林是我国保存相对完好的天然林,也是长江上游重要的生态屏障,在调节区域气候、涵养水源及维持生物多样性等方面有着不可或缺的作用。快速准确估算川西亚高山针叶林的生物量,有助于定量反映亚高山针叶林的生产力,揭示其生态环境健康状况。本研究以岷江源地区内一典型亚高山针叶林区为研究区(地理位置:103.71°-103.73°E,33.03°-33.04°N),以无人机倾斜摄影图像为遥感数据,获取高密度三维点云及高精度数字表面模型(Digital surface model,DSM)和数字高程模型(Digital elevation model,DEM),基于此提取了单木的树冠和树高。在研究区内布设5个样方,并编号为样方1—5,输入样方内每棵单木的实测树高和胸径数据到已有的适用于研究区的树高、胸径—生物量估测方程,估测样方的森林地上生物量,回归分析确定基于树高的地上生物量估测模型。输入单木树高到模型中,实现亚高山针叶林地上生物量的估测。论文主要内容及结论如下:(1)基于无人机可见光影像和二维图像匹配技术,获取的亚高山针叶林三维点云的构建精度高,能准确表达单木。在获取研究区亚高山针叶林无人机可见光影像后,对影像进行空三处理,生成研究区数字正射影像(Digital orthophoto map,DOM)。采用二维图像匹配技术中的运动恢复结构算法(Structer from Motion,SfM)获取森林的三维点云数据。点云数据中植被和地面被明显分离,可清晰辨别单木,点云构建效果较好,验证了基于无人机影像匹配获取亚高山针叶林表面三维点云的能力。(2)提取点云后,点云的滤波和分类是数据处理的关键技术,也是后续信息提取的基础。采用逐渐加密算法,通过反复建立地表三角网模型的方式提取DEM。通过对比点之间高程差的一致性检验DEM精度,实测高程差与对应DEM高程差决定系数R2达0.95。由点云直接生成DSM,采用“Classify by height from ground”算法剔除灌木和草本植物等低矮植被,提取出仅代表亚高山针叶林表面的DSM。提取地面点是提取单木树高的关键,采用无人机可见光影像和点云滤波算法,能有效提取地面点且能保证精度,在亚高山针叶林林区非常适用。(3)利用无人机可见光遥感影像提取亚高山针叶林单木树高,提取精度为0.94,整体上提取值小于实测值。采用面向对象的方法提取单木树冠,总体精度达0.92。DSM与DEM之差即为冠层高度模型(Canopy Height Model,CHM)。叠加单木树冠和CHM,单木范围内的CHM最大值即为单木树高。样方1-4中树高提取值与实测值的决定系数R2达0.94。以26株检验样本反演树高,树高实测值和反演值的R2达0.94,平均反演精度达0.88,满足生物量估测的精度要求。(4)研究区亚高山针叶林基于树高的生物量估测模型为幂函数模型:W=0.0374*H3.4682。通过文献调研,找出适用于研究区内基于树高和胸径的异速生长方程,以实测树高、胸径为输入数据,计算得到样方生物量。对样方生物量和树高进行回归建模,对比常用的四种形式:幂函数模型、线性模型、多项式模型以及指数模型,发现幂函数拟合效果最好,R2达0.98。以提取的单木树高为输入数据,计算得到研究区内所有单木生物量。综上所述,本研究采用SfM算法、点云滤波中的逐渐加密算法,结合面向对象、影像分割和空间分析等技术,探讨了无人机可见光影像在亚高山针叶林DOM、DSM、DEM、CHM、单木树冠以及单木树高的获取方法。其中,单木树冠、树高的提取精度均达到90%以上,单木树高与生物量的R2达0.98。总的来说,基于无人机倾斜摄影的亚高山针叶林生物量估测方法高效可靠,能够满足其生长状况快速评价与动态遥感监测的需求,也为其他森林信息提取提供了思路。
Other AbstractForest biomass refers to the amount of material accumulated by plant per unit area and its essence is organic matter and energy obtained through photosynthesis conducted by assimilating organ of green plant. Western Sichuan subalpine coniferous forest is the relatively unharmed natural forest in China, which is not only the significant timber forest but also the indispensable ecological barrier in upper Yangtze River, and it has an irreplaceable role in regional climate regulation, water conservation and biodiversity maintenance. Rapid and accurate estimation of western Sichuan subalpine coniferous forest biomass can be helpful to reflect its productivity and ecological environment health status.In this study, one typical subalpine coniferous forest in origin area of the Minjiang River was taken as test area (The geographical position: 103.71°-103.73°E,33.03°-33.04°N ). Based on high density three-dimensional point cloud data, high precision DSM (Digital surface model) and DEM (Digital elevation model) produced from UAV slope photography images, individual tree crown and height were extracted. Quadrat forest biomass were calculated using allometric equation based on measured tree height and diameter at breast height. With a regression analysis of the quadrat biomass and tree height, biomass estimation model based on tree height was determined and above-ground biomass of subalpine coniferous forest was estimated.The main research contents and conclusions are as follows:(1) Surface three-dimensional point cloud data of subalpine coniferous forest based on Structer from Motion algorithm and UAV visible images has high precise, and individual tree can be expressed exactly.After obtaining UAV images of subalpine coniferous forest in test area, Digital orthophoto model (DOM) were produced using AgiSoft PhotoScan. Three-dimensional point cloud data were extracted by using Structure from Motion (SfM) algorithm, which verified the ability of UAV images matching to generate subalpine coniferous forest three-dimensional point cloud data.(2) Point cloud filting and classification are the key technologies of data procesing and the base of follow-up information extraction.Firstly reviewed literature on the point cloud filtering algorithm. DEM was extracted based on gradually encryption algorithm by building surface triangulation model. To verify the precision of DEM, consistency of elevation difference between points was compared. As a result, the coefficient of determination of measured elevation difference and corresponding elevation difference on DEM reached 0.95. DSM can be generated from point cloud directly. Low vegetation such as shrub and herbage were rejected using “Classify by height from ground” algorithm to obtain DSM representing pure surface of subalpine coniferous forest.(3) Extraction accuracy of individual tree height is 0.94 from UAV visible images, and extraction values are lower than measured values as a whole.Individual tree crowns were extracted by adopting object-oriented method with 0.92 overall accuracy. The maximum CHM within individual tree is height of the tree after overlaying individual tree crowns and CHM. R2 of measured tree height and extracted tree height in quadrat 1-4 reached 0.94. 26 test tree samples were used to invert tree height, and the R2 reached 0.94, when the average precision was 0.88, which can meet the accuracy requirements of biomass estimation.(4) The best model for estimating subalpine coniferous forest biomass in study area is power function : W = 0.0374*H3.4682.Allometric equation based on tree height and DBH suitable for the study area was found through literature researches. Quadrat biomass was calculated combining measured height and DBH. The regression model was built by comparing three common forms: power function model, linear model, and polynomial model, as a result, the power function fitted best with 0.98 R2. Biomass of individual tree was calculated by inputting extracted tree height data to the power function model.In conclusion, this paper discussed DOM, DSM, DEM, CHM, individual tree crown and height extraction of subalpine coniferous forest based on UAV visible images by employing SfM algorithm and gradually encryption algorithm for point cloud filtering algorithm combining technologies like object-oriented, image segmentation and spatial analysis. Extraction accuracy of individual tree crown and height reached up to more than 0.9, and coefficient of determination of height and biomass reached 0.98. On the whole, biomass estimation of subalpine coniferous forest based on UAV remote sensing images is efficient and reliable and can meet demand of rapid evaluation and dynamic remote sensing monitoring, which also provide ideas for other forest information extraction. 
Pages95
Language中文
Document Type学位论文
Identifierhttp://ir.imde.ac.cn/handle/131551/24755
Collection中国科学院水利部成都山地灾害与环境研究所
Affiliation中国科学院成都山地灾害与环境研究所
First Author Affilication中国科学院水利部成都山地灾害与环境研究所
Recommended Citation
GB/T 7714
王枚梅. 基于无人机倾斜摄影的亚高山针叶林地上生物量估测方法研究[D]. 北京. 中国科学院大学,2018.
Files in This Item:
File Name/Size DocType Version Access License
基于无人机倾斜摄影的亚高山针叶林地上生物(5865KB)学位论文 开放获取CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[王枚梅]'s Articles
Baidu academic
Similar articles in Baidu academic
[王枚梅]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[王枚梅]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 基于无人机倾斜摄影的亚高山针叶林地上生物量估测方法研究.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.