摘要
本次试验以湖南省湘潭县为研究区,提取Landsat 8 OLI影像数据的56个遥感因子作为候选因子,结合皮尔逊相关系数和主成分分析两种方法对变量进行降维,构建多元线性回归模型(MLR)、误差反向传播神经网络(BP-ANN)、K最近邻模型(KNN)和随机森林模型(RF)进行蓄积量反演,并采用决定系数(R^(2))、均方根误差(RMSE)以及相对均方根误差(RRMSE)三个指标对模型进行精度评价。结果表明:三种机器学习模型的拟合结果均优于多元线性回归模型,其决定系数(R^(2))均大于0.6,其中RF最高,为0.67;四种模型中,三种机器学习模型的估测精度均比传统线性模型高出10%以上,其中随机森林模型(RF)精度最高,其均方根误差为57.5 m 3·hm^(-2),相对均方根误差为24.2%。
In this work,Xiangtan County in Hunan Province was collected as the research area.Landsat8 OLI image data was utilized as the remote sensing data source,in which fifty-sixth remote sensing factors were extracted as candidate factors.The Pearson correlation coefficient and principal component analysis were used to reduce the dimensionality of variables.The multiple linear regression model(MLR),error back propagation neural network(BP-ANN),K nearest neighbor model(KNN)and random forest model(RF)were established to inverse volume accumulation.Three coefficients of determination coefficient(R^(2)),root mean square error(RMSE),relative root mean square error(RRMSE%)were used for accuracy evaluation.The results showed that the fitting results of the three machine learning models were better than the multiple linear regression model.Their determination coefficients(R^(2))were all higher than 0.6,in which the highest RF was 0.65.Among the four models,the estimation accuracy of the three machine learning models was higher than the traditional linear model by more than 10 percentage points,of which the random forest model(RF)has the highest accuracy with root mean square error of 66.7 m^(3)·hm^(-2) and the relative root mean square error of 32.3%.
作者
钟健
郑秋斌
ZHONG Jian;ZHENG Qiubin(Guangdong Lingnan Comprehensive Survey and Design Institute,Guangzhou 510000,Guangdong,China)
出处
《湖南林业科技》
2021年第1期61-65,共5页
Hunan Forestry Science & Technology
关键词
遥感影像
森林蓄积量
皮尔逊相关系数
主成分分析
机器学习模型
remote sensing image
forest stock
Pearson correlation coefficient
principal component analysis
machine learning
