基于哑变量的高山松蓄积量反演模型研究  

Application of Dummy Variable in the Research of Pinus Densata Stock Volume Inversion Model

作　　者：王宗梅[1] 徐天蜀[1] 岳彩荣[1] 刘琦[1] 

机构地区：[1]西南林业大学林学院,昆明650224

出　　处：《林业资源管理》2017年第4期75-81,共7页Forest Resources Management

基　　金：国家自然科学基金项目(31260156);西南林业大学科技创新基金项目(C16022)

摘　　要：基于Landsat TM和地面实测样地数据,采用传统线性回归和引入哑变量的线性回归两种建模方法构建香格里拉高山松蓄积量反演模型,并对模型进行验证。研究表明,传统一元和多元线性回归模型的相关系数分别为0.280和0.365,引入哑变量的线性回归模型相关系数为0.602;结合实测检验数据,传统一元、多元线性模型和引入哑变量的模型预测精度分别为61.1%,74.9%和80.3%,引入哑变量的高山松森林蓄积量模型反演精度明显提高,研究结果可为今后基于哑变量的遥感森林蓄积量反演提供一定的依据和参考。Based on Landsat TM and field survey data,two strategies were adopted to construct shangri-la Pinus Densata stock volume inversion model ： conventional linear regression model and linear regression model with dummy variable,and the inversion model was validated. According to the research, correlation coefficients of conventional linear regression with simple regression and multiple regression were 0. 280 and 0. 365 respectively, while the linear regression model with dummy variable had a correlation coeffi-cient of 0. 602 ; Comparing with test sample data, the prediction accuracies of conventional linear regres-sion model were 61. 1% and 74. 9% respectively, while the accuracy of linear regression model with dum-my variable was 80. 3%. It was proven that applying dummy variable could certainly raise the prediction accuracy and provide a reliable reference for forest stock volume inversion via dummy variable in remote sensing to some extent.

关 键 词：哑变量 香格里拉高山松 森林蓄积量 线性回归模型中 

分 类 号：S757[农业科学—森林经理学] TP79[农业科学—林学]