| 运用融合纹理和机载LiDAR特征模型估测森林地上生物量 |
| |
| 引用本文： | 胡凯龙,刘清旺,李世明,庞勇,李梅.运用融合纹理和机载LiDAR特征模型估测森林地上生物量[J].东北林业大学学报,2018(1):52-57. |
| |
| 作者姓名： | 胡凯龙 刘清旺 李世明 庞勇 李梅 |
| |
| 作者单位： | 中国矿业大学(北京) ,北京,100083
中国林业科学研究院资源信息研究所
中国林业科学研究院荒漠化研究所 |
| |
| 基金项目： | 国家自然科学基金项目，国家重点基础研究发展计划项目 |
| |
| 摘 要： | 针对区域尺度森林地上生物量的分布情况,以大兴安岭生态观测站为例,提出了一种融合光学影像纹理和机载LiDAR点云特征的森林地上生物量遥感估测方法。该方法首先提取Landsat 8 OLI不同波段在不同运算窗口下的纹理特征;然后对机载LiDAR点云进行滤波提取地面点,并利用地面点对点云数据进行高度归一化处理,提取点云特征因子;最后结合提取的遥感特征因子,利用支持向量回归的方法对研究区森林地上生物量进行估测,并对结果进行精度验证。结果表明:不同波段和窗口尺寸的建模精度差异较大,蓝光波段在7×7运算窗口下模型精度最高(R~2=0.73,R_(MSE)=22.32 t/hm~2);点云高度分位数变量的建模精度呈正态分布,变量H_(50)的建模精度最高(R~2=0.75,R_(MSE)=19.24 t/hm~2);与单一的遥感特征变量相比,融合光学影像纹理和机载LiDAR点云特征的模型精度有了一定提高,且针叶林和混交林的估测R_(MSE)分别为19.63和20.40 t/hm~2。因此,该方法可以为区域性的森林地上生物量估测提供有效参考。
|
| 关 键 词： | 光学影像纹理 机载激光雷达 森林地上生物量 高度归一化 支持向量回归 Texture of optical image Airborne LiDAR Forest aboveground biomass Height normalization Support vector regression |
Estimation of Forest Aboveground Biomass by Fusion of Optical Image Texture and Airborne LiDAR Metrics |
| |
| Hu Kailong,Liu Qingwang,Li Shiming,Pang Yong,Li Mei.Estimation of Forest Aboveground Biomass by Fusion of Optical Image Texture and Airborne LiDAR Metrics[J].Journal of Northeast Forestry University,2018(1):52-57. |
| |
| Authors: | Hu Kailong Liu Qingwang Li Shiming Pang Yong Li Mei |
| |
| Abstract: | Aiming at the distribution of forest aboveground biomass in regional scale, a method of estimating forest aboveground biomass based on the optical image texture and airborne LiDAR metrics was proposed in the Daxing' an Mountains forest e-cosystem station, and the influence of different feature factors was analyzed on the estimation accuracy.Firstly, the texture features of different bands of Landsat 8 OLI were extracted.Secondly, the airborne LiDAR point cloud was filtered to ex-tract the ground points.The point cloud was normalized by ground points, and LiDAR metrics were extracted.Finally, the forest aboveground biomass was estimated by the extracted remote sensing feature factor by support vector regression, and the accuracy of the results was verified by in-situ data.The modeling accuracy of different bands and window sizes were quite different.The model accuracy was highest in blue band with the 7×7 window size, R2 of 0.73, and RMSE of 22.32 t/ha.The model accuracy of the LiDAR metrics was normal distribution, and the highest accuracy of variable H50 was with R2=0.75 and RMSE of 19.24 t/ha.Compared with the single variable, the model accuracy of the fusion optical image tex-ture and the airborne LiDAR metrics was improved, and the estimated RMSE of coniferous and mixed forest were 19.63 and 20.40 t/ha, respectively. |
| |
| Keywords: | |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
