Estimation of Nitrogen Content in Ramie Leaves Based on Multiscale Hyperspectral Technology
Chen Jianfu
Yue Yunkai
Fu Hongyu
Xu Mingzhi
Jiao Xinwei
Liao Ao
Zhang Lei
Zhao Liang
Chen Zhaozhong
Cui Guoxian
She Wei
Abstract：Nitrogen fertilizer is important for ramie growth and yield and quality formation.Traditional ni-trogen detection methods are time-consuming and labor-intensive,and have poor timeliness,etc.Hyperspectral remote sensing technology can effectively monitor a number of physiological indicators of crops,thus help to grasp crop growth and nutritional status in real time.In this study,we set four nitrogen application levels,i.e.,N0(no fertilization),N1(273 kg/hm2 pure nitrogen),N2(332 kg/hm2 pure nitrogen),N3(390 kg/hm2 pure nitrogen,conventional application rate),and two stages of fertilizer application,i.e.,the row closure stage(a)and the vigorous stage(b),and then used spectral technique to analyze the differences of spectral characteristics under different treatments at two scales of crown and leaves.By analyzing the correla-tion between multi-scale hyperspectral data and nitrogen content in leaves(LNC)of ramie,three machine learning methods,namely partial least squares regression(PLSR),support vector machine(SVM)and ran-dom forest(RF)were used to construct the models for estimation of ramie LNC at key growth stages(seedling stage,row closure stage,vigorous stage and maturity stage),and then the comparative analyses were carried out.The results showed that the accuracy of the canopy spectral model was better than that of the leaf spectral model at several growth stages.The accuracy of the ramie LNC estimation model based on canopy spectral band was the highest at maturity stage(R2=0.795,RMSE=0.608),and that based on leaf spectral band was the highest at vigorous stage(R2=0.670,RMSE=0.470).The LNC estimation model constructed based on RF algorithm was the best in multiple growth periods with stable performance and higher accuracy.In conclusion,the estimation of ramie leaf nitrogen content based on multi scale hyperspectral technology was feasible,espe-cially at the canopy scale.This study provided a better means for ramie nitrogen inversion.
Keywords：RamieLeaf nitrogen contentHyperspectralMultiscale
Publication Date：2025-02-28
Online Publishing Date：2026-05-22(First online date of this platform, not the publication date of the document)
Pages：7( 165-171 )