基于随机森林模型的林地叶面积指数遥感估算
发布时间：2018-10-05 14:13
【摘要】：林地叶面积指数(Leaf area index,LAI)的准确估测是精准林业的重要体现。为了快速、准确、无损监测林地LAI,利用LAI-2200型植物冠层分析仪获取福建省西部森林样地的LAI数据,结合同期Pleiades卫星影像计算12种遥感植被指数,分析了各样地实测LAI数据和相应植被指数的相关性,进而使用随机森林(RF)算法构建了林地LAI估算模型,以支持向量回归(SVR)模型和反向传播神经网络(BP)模型作为参比模型,以决定系数(R~2)、均方根误差(RMSE)、平均相对误差(MAE)和相对分析误差(RPD)为指标评价并比较了模型预测精度。结果表明:全样本数据中,各植被指数与对应LAI值均呈极显著相关(P0.01),且相关系数都大于0.4;RF模型在3次不同样本组中的预测精度均高于同期的SVR模型和BP模型;3个样本组中RF模型的LAI估测值与实测值的R~2分别为0.688、0.796和0.707,RPD分别为1.653、1.984和1.731,均高于同期SVR模型和BP模型,对应的RMSE分别为0.509、0.658和0.696,MAE分别为0.417、0.414和0.466,均低于同期其他2种模型。
[Abstract]:The accurate estimation of forest leaf area index (Leaf area index,LAI) is an important embodiment of precision forestry. In order to quickly, accurately and accurately monitor forest land LAI, to obtain LAI data of forest sample land in western Fujian Province by LAI-2200 plant canopy analyzer, and to calculate 12 remote sensing vegetation indices by combining Pleiades satellite images. The correlation between the measured LAI data and the vegetation index was analyzed, and then the estimation model of forest land LAI was constructed by using the stochastic forest (RF) algorithm. The support vector regression (SVR) model and the backpropagation neural network (BP) model were used as the reference model. The prediction accuracy of the model was evaluated and compared with the determination coefficient (Ru 2), the mean relative error (MAE) of the root mean square error (RMSE),) and the relative analysis error (RPD). The results show that, in the whole sample data, Each vegetation index was significantly correlated with the corresponding LAI value (P0.01), and the correlation coefficient was greater than 0.4g RF model. The prediction accuracy of RF model in three different sample groups was higher than that of SVR model and BP model in the same period, and the LAI estimation value of RF model in three sample groups was higher than that of RF model. Compared with the measured values, the RPD-values of RN-2 were 0.688U 0.796 and 0.707 g, respectively, which were 1.653U 1.984 and 1.731respectively, which were higher than those of SVR model and BP model. The corresponding RMSE were 0.509 and 0.658, respectively, and 0.669 were 0.417, 0.414 and 0.466, respectively, which were lower than those of the other two models at the same time.
【作者单位】： 福建农林大学3S技术与资源优化利用福建省高校重点实验室;福建农林大学林学院;
【基金】：国家自然科学基金项目(41401385)
【分类号】：S771.8;TP79
本文编号：2253708
[Abstract]:The accurate estimation of forest leaf area index (Leaf area index,LAI) is an important embodiment of precision forestry. In order to quickly, accurately and accurately monitor forest land LAI, to obtain LAI data of forest sample land in western Fujian Province by LAI-2200 plant canopy analyzer, and to calculate 12 remote sensing vegetation indices by combining Pleiades satellite images. The correlation between the measured LAI data and the vegetation index was analyzed, and then the estimation model of forest land LAI was constructed by using the stochastic forest (RF) algorithm. The support vector regression (SVR) model and the backpropagation neural network (BP) model were used as the reference model. The prediction accuracy of the model was evaluated and compared with the determination coefficient (Ru 2), the mean relative error (MAE) of the root mean square error (RMSE),) and the relative analysis error (RPD). The results show that, in the whole sample data, Each vegetation index was significantly correlated with the corresponding LAI value (P0.01), and the correlation coefficient was greater than 0.4g RF model. The prediction accuracy of RF model in three different sample groups was higher than that of SVR model and BP model in the same period, and the LAI estimation value of RF model in three sample groups was higher than that of RF model. Compared with the measured values, the RPD-values of RN-2 were 0.688U 0.796 and 0.707 g, respectively, which were 1.653U 1.984 and 1.731respectively, which were higher than those of SVR model and BP model. The corresponding RMSE were 0.509 and 0.658, respectively, and 0.669 were 0.417, 0.414 and 0.466, respectively, which were lower than those of the other two models at the same time.
【作者单位】： 福建农林大学3S技术与资源优化利用福建省高校重点实验室;福建农林大学林学院;
【基金】：国家自然科学基金项目(41401385)
【分类号】：S771.8;TP79
【相似文献】
相关博士学位论文 前1条
1 雷震;随机森林及其在遥感影像处理中应用研究[D];上海交通大学;2012年
相关硕士学位论文 前1条
1 姚明煌;随机森林及其在遥感图像分类中的应用[D];华侨大学;2014年
，本文编号：2253708
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2253708.html
