基于哑变量和因子选择的森林蓄积量估测研究

Forest Volume Estimation Based on Dummy Variables and Factor Selection

为提高森林蓄积量遥感估测精度,探讨哑变量技术在蓄积量遥感估测中的作用。以云南省普洱市思茅区为研究区,以Landsat 8 OLI和93块森林资源二类调查角规控制样地数据为基础,使用随机森林(random forest)算法进行遥感变量因子的选择,并以龄组为哑变量分别构建基于哑变量的SVR和PLSR蓄积量估测模型,采用留一交叉验证对结果进行评估。结果表明,使用随机森林算法进行变量的选择有效减少了自变量的维度,提高了计算效率;其次,哑变量引入后,PLSR和SVR 2种回归模型的估测精度都比无哑变量方法有明显的提高,且SVR的估测结果优于PLSR;在引入哑变量后SVR模型的决定系数R^(2)由0.59提高到0.68,相对均方根误差rRMSE由36.76%降低至32.97%,PLSR模型的决定系数R^(2)由0.53提高到0.62,相对均方根误差rRMSE由39.41%降低至35.24%。在森林蓄积量的遥感估测中,哑变量技术的应用可以在一定程度上解决不同蓄积量大小对估测结果造成的影响,进而提高蓄积量的估测精度。

Forest volume is one of the important forest parameters.In order to improve the accuracy of remote sensing estimation of forest volume,the role of dummy variable technology in forest volume estimation was discussed.Taking Simao District,Pu'er City,Yunnan Province as the study area,based on the data of Landsat8 OLI and survey data of 93 sample plots,the remote sensing variable factors were selected by using random forest algorithm,and the support vector regression(SVR)and partial least squares regression(PLSR)estimation models based on dummy variables were constructed,respectively with age groups as dummy variables,and the results were evaluated by Leave one cross-validation.Firstly,the results showed that using random forest algorithm to select variable effectively reduced the dimension of independent variable and improved the calculation efficiency.Secondly,after the introduction of dummy variables,the estimation accuracies of PLSR and SVR regression models were significantly improved than the method without dummy variables,and the estimation result of SVR was better than PLSR.After introducing the dummy variable,the decision coefficient R^(2) of the SVR model increased from 0.59 to 0.68,the relative root mean square error rRMSE decreased from 36.76%to 32.97%,the decision coefficient R^(2) of the PLSR model increased from 0.53 to 0.62,and the relative root mean square error rRMSE decreased from 39.41%to 35.24%.In the remote sensing estimation of forest stock volume,the application of dummy variable technology can solve the influence of different stock volume on the estimation results to a certain extent,and then improve the estimation accuracy of stock volume.