| 运用GF-1影像光谱和纹理信息构建森林蓄积量估测模型 |
| |
| 引用本文： | 刘伯涛,李崇贵,郭瑞霞,刘思涵,马婷.运用GF-1影像光谱和纹理信息构建森林蓄积量估测模型[J].东北林业大学学报,2020(1):9-12,28. |
| |
| 作者姓名： | 刘伯涛 李崇贵 郭瑞霞 刘思涵 马婷 |
| |
| 作者单位： | 西安科技大学 |
| |
| 基金项目： | “十三五”国家重点研发计划(2017YFD0600400) |
| |
| 摘 要： | 以GF-1遥感影像为数据源,研究区森林资源二类调查数据为样地实测数据,综合考虑光谱、地形、纹理特征,利用多元线性回归、BP神经网络、支持向量机和随机森林建立研究区森林蓄积量估测模型,并验证模型预测的性能。结果表明:4种模型预测评价指标的决定系数(R2)和均方根误差(RMSE)相近,但有一定的差异,多元线性回归模型R2和RMSE分别为0.446、39.979 6 m^3·hm^-2,BP神经网络模型R2和RMSE分别为0.474、39.703 9 m^3·hm^-2,支持向量机模型R2和RMSE分别为0.485、38.924 8 m^3·hm^-2,随机森林模型R2和RMSE分别为0.534、37.882 2 m^3·hm^-2;3种机器学习方法构建的蓄积量估测模型预测性能优于传统的多元线性回归模型,随机森林模型的预测性能最优。
|
| 关 键 词： | GF-1遥感影像 森林蓄积量 多元线性回归 随机森林 支持向量机 BP神经网络 |
Building Forest Volume Estimation Model Using GF-1 Image Spectral and Texture Information |
| |
| Liu Botao,Li Chonggui,Guo Ruixia,Liu Sihan,Ma Ting.Building Forest Volume Estimation Model Using GF-1 Image Spectral and Texture Information[J].Journal of Northeast Forestry University,2020(1):9-12,28. |
| |
| Authors: | Liu Botao Li Chonggui Guo Ruixia Liu Sihan Ma Ting |
| |
| Institution: | (Xi’an University of Science and Technology,Xi’an 710054,P.R.China) |
| |
| Abstract: | Taking the GF-1 remote sensing image as the data source and the second-class survey data of forest resources in the study area as the field survey data, the forest accumulation estimation model was established by taking the spectral, topographic and texture characteristics into comprehensive consideration, using the multiple linear regression, BP neural network, support vector machine and random forest. The prediction performance of the model was verified. The coefficient of determination(R2) and root mean square error(RMSE) of the four model prediction evaluation indexes are as follows: those of the multiple linear regression models are 0.446 and 39.979 6 m^3·hm^-2, those of the BP neural network models are 0.474 and 39.703 9 m^3·hm^-2, those of the SVM models are 0.485 and 38.924 8 m^3·hm^-2, and those of the random forest models are 0.534 and 37.882 2 m^3·hm^-2, respectively. The prediction performance of the forest accumulation estimation model constructed by three machine learning methods are better than that by the traditional multiple linear regression model, and the random forest model has the best prediction performance. |
| |
| Keywords: | GF-1 remote sensing image Forest accumulation Multiple linear regression Random forest Support vector machines BP neural network |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
