软件导刊2025，Vol.24Issue(5)：62-69,8.DOI:10.11907/rjdk.241132
基于机器学习的不同生育期下玉米叶片叶绿素含量反演研究
Research on the Inversion of Chlorophyll Content in Maize Leaves at Different Growth Stages Based on Machine Learning
摘要
Abstract
Chlorophyll,as an important pigment for ensuring crop yield and quality,timely and accurate monitoring of crop chlorophyll con-tent is of great significance for achieving precision agriculture.This study focuses on corn as the research object,using unmanned aerial vehi-cle multispectral remote sensing to collect spectral data of three important growth stages(seedling stage,jointing stage,and maturity stage)of corn,and synchronously collecting chlorophyll content of corn leaves.Reflectivity is extracted from multispectral images,and spectral indices are calculated.The optimal combination of independent variables for spectral indices at different growth stages is selected through grey correla-tion method as the input variable of the model.Based on this,three extreme learning machines(ELM),random forest(RF),and BP propaga-tion neural network(BPNN)chlorophyll content monitoring models are constructed for different growth stages.Evaluate the accuracy of the constructed monitoring model.The results indicate that different growth stages and different models have a significant impact on monitoring ac-curacy.In terms of reproductive period,all three models used in this study achieved the best monitoring accuracy during the jointing stage;In terms of the model,random forests achieved high accuracy in all three growth stages,with the accuracy ranking as ELM(R2=0.57-0.65;RMSE=1.32-1.44)<BPNN(R2=0.56-0.68;RMSE=1.23-1.45)<RF(R2=0.61-0.71;RMSE=1.04-1.23);From this,it can be concluded that the grey correlation method ranks spectral indices differently at different growth stages.All three models achieved the highest accuracy dur-ing the jointing stage,with similar accuracy during the seedling and mature stages;During the jointing stage,RF has the highest accuracy and the best model stability.关键词
多光谱影像/灰色关联法/机器学习/精准农业Key words
multispectral image/grey relational analysis/machine learning/precision agriculture分类
信息技术与安全科学引用本文复制引用
安学兰,聂志刚,李广,鄢继选..基于机器学习的不同生育期下玉米叶片叶绿素含量反演研究[J].软件导刊,2025,24(5):62-69,8.基金项目
甘肃农业大学青年导师扶持基金项目(GAU-QDFC-2022-19) （GAU-QDFC-2022-19）
甘肃省教育厅产业支撑计划项目(2021CYZC-15,2022CYZC-41) （2021CYZC-15,2022CYZC-41）
