Position: Home > Articles > Hyperspectral Estimation of Biomass of Winter Wheat at Different Growth Stages Based on Machine Learning Algorithms
Journal of Triticeae Crops
2019
(2)
217-224
基于机器学习算法的冬小麦不同生育时期生物量高光谱估算
作 者：
吴芳;李映雪;张缘园;张雪红;邹晓晨
单 位：
南京信息工程大学应用气象学院
关键词：
冬小麦生物量;高光谱估算;随机森林算法;支持向量回归;偏最小二乘算法
摘 要：
为探索适用于冬小麦不同生育时期的高光谱估算方法,基于4年大田试验,以江苏省主要冬小麦品种为材料,以8种对常用生物量敏感的高光谱指数为基础,分别采用偏最小二乘算法、支持向量回归算法、随机森林算法在冬小麦4个主要生育时期(抽穗期前、抽穗期、开花期和灌浆期)进行了高光谱生物量估算和预测能力比较。结果表明,在冬小麦不同生育时期,高光谱估算生物量精度差异显著;利用随机森林构建的生物量估算模型在4个生育时期均表现出很好的效果,决定系数(r~2)和均方根误差(RMSE)在抽穗期前分别为0.79和44.82 g·m~(-2),在抽穗期分别为0.71和62.07 g·m~(-2),在开花期分别为0.70和97.63 g·m~(-2),在灌浆期分别为0.71和106.98 g·m~(-2);随机森林模型在4个生育时期的预测能力都高于或接近于支持向量回归模型,高于偏最小二乘回归模型,r~2和RMSE在抽穗期前分别为0.60和72.54 g·m~(-2),在抽穗期分别为0.60和75.07 g·m~(-2),在开花期分别为0.68和109.9 g·m~(-2),在灌浆期分别为0.61和127.93 g·m~(-2)。这说明随机森林算法在冬小麦不同生育时期生物量高光谱遥感估算方面具有较高的精度和稳定性。
译 名：
Hyperspectral Estimation of Biomass of Winter Wheat at Different Growth Stages Based on Machine Learning Algorithms
作 者：
WU Fang;LI Yingxue;ZHANG Yuanyuan;ZHANG Xuehong;ZOU Xiaochen;Nanjing University of Information Science and Technology,College of Applied Meteorology;Nanjing University of Information Science and Technology,School of Remote Sensing and Geomatics Engineering;
单 位：
WU Fang%LI Yingxue%ZHANG Yuanyuan%ZHANG Xuehong%ZOU Xiaochen%Nanjing University of Information Science and Technology,College of Applied Meteorology%Nanjing University of Information Science and Technology,School of Remote Sensing and Geomatics Engineering
关键词：
Winter wheat biomass;;Hyperspectral estimation;;Random forest algorithm;;Support vector regression;;Partial least squares algorithm
摘 要：
In order to explore the optimal hyperspectral estimation method at different growth stages of winter wheat,the experiments were carried out based on a 4-year field test.With the wheat cultivars of Jiangsu province as materials,eight kinds of commonly used biomass-sensitive vegetation indices were used as model inputs.Three machine learning algorithms[partial least squares(PLS),support vector regression,and random forest] were constructed to predict winter wheat biomass at the four main growth stages(before heading stage,heading stage,flowering stage,and grain filling stage).The results showed that the effect of growth stages on winter wheat prediction was significant using hyperspectral remote sensing; among the three algorithms,random forest performed better than the other two algorithms at all four growth stages; the coefficient of determination r~2 and root mean square error RMSE were 0.79 and 44.82 g·m~(-2) before heading date,0.71 and 62.07 g·m~(-2) at heading date,0.70 and 97.63 g·m~(-2)at flowering stage,and 0.71 and 106.98 g·m~(-2) at grain filling stage,respectively.The predictability of the random forest model at the four growth stages was higher or close to that of the support vector regression model,which was higher than that of the PLS-biomass model.The r~2 and RMSE were 0.60 and 72.54 g·m~(-2) before heading stage,0.60 and 75.07 g·m~(-2) at heading stage,0.68 and 109.9 g·m~(-2) at flowering period,and 0.61 and 127.93 g·m~(-2) at grain filling stage,respectively.This result showed that random forest algorithm performed relative high accuracy and robustness for estimating winter wheat biomass at various growth stages using hyperspectral remote sensing.
导出引用
计量
文章访问数: 4
HTML全文浏览量: 0
PDF下载量: 0
所属期刊
