基于高光谱的弼猴桃叶片氮磷钾元素快速检测方法研究

打开文本图片集

中图分类号：S24；S663.4 文献标识码：A 文章编号：2095-5553（2026）04-0195-11

Abstract：Kiwifruit tres are highlysusceptibletonutrient deficiencies，whichcanaffecttree healthandfruit yield.To rapidlydetectthecontentof nitrogen，phosphorus and potassium elements in kiwifruit tresandmonitorthehealth status ofthetrees，takingtheleavesof thekiwifruittreecanopyastheresearchobject，11Okiwifruitleaveswerecollcted from thecanopy layer during the maingrowth stages，hyperspectral images of each sample were acquired byusinga hyperspectral imager with aspectralrangeof 865.11—1715.O3 nm，andthecontentsof elementssuchas nitrogen， phosphorus and potassium in the leaves were measured by using chemical methods forconstructing a hyperspectral data set of canopy leaves.The spectral data were pre-processed respectively byusing derivative algorithms（F—D，S—D）， multiplicative scatercorrction（MSC）andstandard normalvariable transformation（SNV）．Principalcomponentanalysis （PCA）and competitive adaptive re-weighted sampling（CARS）were used to select optimal feature bands.Three modelling methodssupport vectorregression（SVR），partial least squaresregression（PLSR）andrandom forest regression （RFR）wereusedtoconstruct models fortheidentificationof nitrogen，phosphorusandpotassumelement contents.Theexperimental results showed thatthe optimal detectionmodel for nitrogen element was theRFR model， with Rp 2 and RMSEP values of O.862 2 and O.179 6，respectively. The optimal detection model for phosphorus was also the RFR model with Rp2 and RMSEP values of 0.900 9 and O.313 4 respectively. The optimal detection model for potassium was the PLSR model with Rp2 and RMSEP values of O.852 1 and O.389 5 respectively. Thus，the intelligent detectionmodel basedonhyperspectralimaging technologycombinedwith machine learning canrapidly detectthe nitrogen，phosphorusand potassumcontent in kiwifruit leaves.This facilitates theanalysisofthe health statusof the trees，providing crucial decision support for the subsequent intelligent management of the orchard.

Keywords：kiwifruit；hyperspectral; leaf N，P，K；characteristic bands；random forest regression

0 引言

数字化是猕猴桃产业高质量发展的关键[1]。（剩余15463字）