农业机械学报2025，Vol.56Issue(9)：566-575,595,11.DOI:10.6041/j.issn.1000-1298.2025.09.047
基于高光谱红边偏度、峰度和机器学习的植被叶绿素含量估算模型研究
Estimating Vegetation Chlorophyll Content Based on Hyperspectral Red Edge Skewness,Kurtosis,and Machine Learning
摘要
Abstract
Chlorophyll is a key component in plant photosynthesis and an important indicator of physiological status and plant health.Accurate estimation of chlorophyll content is essential for monitoring plant growth conditions.Although traditional rededge parameters were widely used in chlorophyll estimation,they presented certain limitations.To address these shortcomings,rededge skewness(SRE)and rededge kurtosis(KRE)were introduced as novel rededge parameters to explore their relationship with chlorophyll content.Taking winter wheat and summer maize as study subjects,three modeling approaches-multiple linear regression(MLR),extreme gradient boosting(XGBoost),and backpropagation neural network(BPNN)—were employed to compare the performance of chlorophyll estimation models by using the proposed parameters versus traditional rededge parameters.The results demonstrated that SRE and KRE were significantly negatively correlated with chlorophyll content,showing a decreasing trend as chlorophyll levels were increased.Across all three modeling methods and crop types,models incorporating SRE and KRE achieved higher prediction accuracy than those based on traditional rededge parameters,characterized by higher R2,lower RMSE,and more concentrated residual distributions,indicating enhanced model stability and precision.Among the three modeling approaches,the BPNN model consistently exhibited the best performance,with significantly higher R2 values compared with that of MLR and XGBoost.The research result confirmed that SRE and KRE provided a more comprehensive representation of chlorophyll variation than traditional rededge parameters.When combined with BPNN modeling,they can substantially improve the accuracy of chlorophyll content estimation.These findings can offer a novel technical pathway for high-precision chlorophyll monitoring and had practical implications for crop condition assessment and precision agriculture.关键词
叶绿素含量/估算模型/红边偏度/红边峰度/机器学习Key words
chlorophyll content/estimation model/red edge skewness/red edge kurtosis/machine learning分类
农业科技引用本文复制引用
姚付启,曾凡超,孙金伟,饶志龙,王子涵..基于高光谱红边偏度、峰度和机器学习的植被叶绿素含量估算模型研究[J].农业机械学报,2025,56(9):566-575,595,11.基金项目
国家自然科学基金项目(51809284、51309016)和山东省自然科学基金项目(ZR2020ME254、ZR2020QD061) （51809284、51309016）
