Research On Monitoring The Nitrogen Nutritional Parameters At Different Growth Stages Of Sugar Beet Based On Hyperspectrum

In this study,the field experiment of sugar beet was conducted at the Chifeng experimental station in 2014 and at the farm of Inner Mongolia Agricultural University in 2015.This research was based on the field plot experiment of sugar beet for two years.Under different nitrogen levels,the research on monitoring the nitrogen content and SPAD value of sugar beet at different growth stages(seedlings stage,leaves stage,root stage and sugar accumulation stage)was carried out based on hyperspectral technology.The main work and research results of this study are as follows:1.The canopy spectral reflectance of sugar beet at different growth stages was obtained by using ASD hyper-spectrometer.The characteristic of canopy hyperspectral response of sugar beet was researched.And it is found that in the area of visible light(450-680 nm)and near-infrared(760-950 nm),the spectral response is significantly different at different growth stages of sugar beet.The correlation between canopy spectra which were pretreated by smoothing and first derivative and nitrogen content(and SPAD value)was analyzed.The results of correlation analysis shows that the first derivative spectra are more sensitive to nitrogen content and SPAD value,and the correlation is greatly improved.2.The optimal spectral parameters were used as input variables to establish hyperspectral prediction models,and the effect of different algorithms(unitary regression,PCR,PLSR,SVM)on the prediction accuracy of the nitrogen nutritional parameters at different growth stages was researched.The results shows that in the seedling stage,the best prediction model for nitrogen content and SPAD value is respectively PLSR model and PCR model,and the R2 is respectively 0.520 and 0.552,RMSE is respectively 2.70g/kg and 1.99,RE is respectively 6.6%and 3.5%;in the leaves stage,SVM model has the best prediction effect on nitrogen content(R2,RMSE and RE is respectively 0.443,3.17g/kg,and 10.1%),and the quadratic function model established by SDr/SDy has the best prediction effect on SPAD value(R2,RMSE and RE is respectively 0.212,1.85 and 3.3%);in the root stage,the best prediction models for nitrogen content and SPAD value are both PCR models,and the R2 is respectively 0.447 and 0.536,RMSE is respectively 2.93g/kg and 1.81,RE is respectively 9.7%and 3.3%;in the sugar accumulation stage,PLSR model has the best prediction effect on nitrogen content(R2,RMSE and RE is respectively 0.344,2.95g/kg and 9.0%),and the logarithmic function model established by SDb has the best prediction effect on SPAD value(R2,RMSE and RE is respectively 0.324,3.17 and 5.9%).In this research,the more accurate hyperspectral prediction models for nitrogen content and SPAD value at different growth stages were established,so as to provide theoretical basis for fertilization management in sugar beet field and provide technical support for rapid and nondestructive monitoring nutritional parameters of sugar beet.