【作者】 吴红波；

【导师】 邢艳秋；

【作者基本信息】 东北林业大学 ， 森林工程， 2011， 硕士

【摘要】 激光雷达LiDAR (Light Detection and Ranging, LiDAR)技术是近年来遥感探测、对地观测、林业调查的热点之一,在国内外森林调查和制图方面有着广泛的应用,主要集中在森林冠层高度和生物量的估测研究。为此,本研究以长白山系的吉林省汪清林业局经营区为研究区,首先回顾了激光雷达技术在林业上的国内外应用现状与动态,分析了其研究问题以及本研究的主要内容和预期目标；然后,针对大光斑激光雷达ICESat-GLAS (the Ice, Cloud, and land Elevation-Geoscience Laser Altimeter System)回波数据预处理及参数提取方法、影响森林冠层高度估测的因素(坡度等)进行了分析与研究。最后,为验证ICESat-GLAS回波波形参数与森林平均冠层高度和生物量关系,结合野外调查数据构建ICESat-GLAS回波参数与森林平均冠层高度、生物量的估测模型,并利用GIS空间分析技术实现LiDAR回波数据高精度地估测森林平均冠层高度和地上生物量。主要研究结果如下：1)采用小波变换和高斯滤波器分别对获取的ICESat-GLAS回波波形数据进行去噪,去噪后的ICESat-GLAS回波波形可以进行提取波形参数。与高斯滤波器相比,经小波变换后的LiDAR回波波形的均方根误差(Root Mean Square Error, RMSE)降低了0.7188,信噪比(Signal to Noise Ratio, SNR)提高了16.17dB；小波变换滤波较好地保留了回波波形中的有用信息,有效地抑制了回波波形的“叠加”问题,去噪效果优于高斯滤波器的平滑滤波。2)本研究利用多元统计回归方法通过分析ICESat-GLAS回波波形参数(Extent、R20、R50等)与样地平均冠层高度、森林生物量的实测值进行相关分析,并根据最大复相关系数求出平均冠层高度、森林生物量估测方程。从拟合结果可知,样地的森林平均冠层高度、森林生物量估测方程的复相关系数分别为0.801、0.710,预估精度分别为82.59%、80.56%。3)不同空间分辨率的DEM数据对森林平均冠层高度估测有一定的影响。77个检验样地的森林平均冠层高度实测值为21.30m,不同分辨率(20m、30m、90m)冠层高度模型CHM的森林平均冠层高度估测值分别为15.6m、18.9m、17.4m,预估精度分别为73.23%、88.73%、81.69%。经移动窗口差分滤波后的CHM模型消除坡度对估测的影响,且空间分辨率为30m时,此时森林平均冠层高度估测值为19.2m,预估精度为90.14%。4)分析样地坡度与森林平均冠层高度的估测值的关系方程式可知：当样地坡度≤10°时,样地坡度对森林平均冠层高度估测影响较小,波形展宽与坡度的相关性较弱；当样地坡度≥30°时,坡度对森林平均冠层高度估测影响较大,回波波形展宽呈现指数几何增长,呈现严重的“重叠”现象。5)利用GIS软件提取空间分辨率为30m的森林冠层高度和森林地上生物量数据,结果表明：研究区的森林平均冠层高度估测值为18.7m。77个检验样地的森林平均冠层高度估测值为19.2m,预估精度为90.14%。剔除农田、建筑用地、公路、河流等非林业用地,研究区的森林平均冠层高度估测值为23.1m。研究区森林平均地上生物量估测值为91.017t/ha,去除非林业用地后的森林平均地上生物量估测值为104.561t/ha,研究区森林地上生物量总量为5,747,996t。以上研究结果可为ICESat-GLAS波形数据处理和基于大光斑激光雷达波形数据对林分平均冠层高度和生物量的估测应用研究提供方法基础。更多还原

【Abstract】 LiDAR (Light Detection and Ranging) is one of the research focuses on remote sensing, Earth observation, forestry investigation in recent years, and forest survey and mapping in China and abroad have a wide range of applications. The recent studies concentrate on the estimation of forest canopy height and forestry biomass. Therefore, the paper take Jilin province’s Wangqing Forestry Bureau as the study area, where is located in the low mountain in the Changbai Mountains. The first chapter reviews of the status and trends on LiDAR technology applications in forestry at home and abroad, and analysis of its research problems, main contents and expected objectives in this study; Then, the paper has analysis of ICESat-GLAS (the Ice, Cloud, and land Elevation-Geoscience Laser Altimeter System) return waveforms preprocessing and extraction parameter method, and also studies on factors (plot slope, etc.) that effect on forest canopy height estimation. Finally, to verify the relationship between ICESat-GLAS return waveform parameters and the mean forest canopy height and forest aboveground biomass, the research combines with field survey data to construct the mean forest canopy height and forest aboveground biomass retrieval model based on ICESat-GLAS return waveform parameters. Meanwhile, the research also uses the GIS spatial analysis techniques to make further improvement on the predicted accuracy on forest mean canopy height and aboveground biomass. The main results are as follows:1) Respectively using Wavelet Transform method and Gaussian Filter to obtain the ICESat-GLAS return waveforms denoising, the paper extracts the denoised return waveform parameters for estimation of the forest mean canopy height and forest biomass. Its results show that compared with the Gaussian Filter, Wavelet Transform makes the Root Mean Square Error (RMSE) of ICESat-GLAS return waveforms decreased 0.7188, Signal to Noise Ratio (SNR) of that increased 16.17 dB; Wavelet Transform betters than Gaussian Filter in retaining the useful information and suppressing "overlay" problem of the ICESat-GLAS return waveforms.2) The research uses multivariate statistical regression method to analysis the correlation relationship between ICESat-GLAS return waveform parameters(Extent、R20、R50 etc.) and the measured plot mean canopy height, measured forest aboveground biomass. And then according to the correlation coefficient of each other, the paper calculates the forest mean canopy height, abovegroud biomass estimation equation. From the fitting results, it shows that these coefficients of multiple correlation are 0.801,0.710 respectively, these predicted accuracy are 82.59%,80.56%respectively.3) The different spatial resolution of Eigital Elevation Model (DEM) data has a certain impact on estimating forest mean canopy height. The forest measured mean canopy height of 77 test plots is 21.30 meters. The estimated mean canopy height from forest Canopy Height Model (CHM) with different resolutions (20m,30m,90m) is 15.6 meters,18.9 meters,17.4 meters, the predicted accuracy is 73.23%,88.73%, respectively. The moving window-difference Filter can eliminate the impact of slope on estimating forest mean canopy height. When the spatial resolution of filtered CHM is 30 meters, the predicted forest mean canopy height is 19.2 meters; and its predicted accuracy is 90.14%.4) By anglysis of the relationship between the plot slope and estimated forest canopy height value, its regression equation shows that:when the plots slope is less than 10°, the effect of plot slope on forest mean canopy height estimation is weak and its effection is ignored. When the plot slope is more than 30°or more, the slope of the forest canopy height estimation is larger and the waveform broadening associated with the plot slope shapes in exponential growth, showing significant "overlap" problems within the return waveforms.5) By GIS software to extract a forest canopy height and aboveground biomass raster data with a spatial resolution of 30m, the results show that the, the estimated forest mean canopy height value of whole study area is 18.7 meters. The measured forest mean canopy height value of 77 test plots is 19.2 meters, its predicted accuracy is 90.14%. Removal of farmland, buildings, roads, rivers and other non-forest land, the estimated forest mean canopy height value of the study area is 23.1 meters. The estimated overall forest aboveground biomass is 91.017 ton/hectare. Removal of non-forestry land, the estimated overall forest aboveground biomass is 104.561 ton/hectare. The total forest aboveground biomass of study area is 5,747, 996 tons.The research results provide a new method of ICESat-GLAS waveforms data processing and also provide scientific references to estimate forest mean canopy height and forest above-ground biomass based on large-footprint waveform data.更多还原