Monitoring Nitrogen Status With Hyperspectral Imaging In Wheat

Quantitative monitoring of plant nitrogen status has become an important research field of remote sensing of vegetation at home and abroad.With rapid,non-destructive,accurate way to estimate crop nitrogen nutrition status is one of the key technologies of precision agriculture development.The purpose of this study is based on wheat as object,based on different years,different nitrogen levels,different planting density and different varieties of field experiment,Using high imaging spectrometer to obtain wheat near ground of hyperspectral images,based on theory and mathematical morphology spectroscopy method using a double coupling spectrum and texture characteristics,the separation the wheat didn’t remove background,wheat remove background,light wheat and shadow wheat,respectively,and then use successive projection algorithm to extract the characteristic band of each target component,using Gray-level Co-occurrence Matrix(GLCM)get texture information of each characteristic band,and then analysis based on the spectral information,texture information,and texture and spectral information to build the best monitoring models,which is a portable monitor crop nitrogen nutrition imaging research,development and use of space remote sensing analytical information provides the core band selection,for the wheat nitrogen nutrition of real-time monitoring and accurate diagnosis to provide effective technical support.Firstly,under fully considering the different growth period of wheat canopy imaging spectral characteristics and the growth condition of the field,Using new spectral index classification,texture classification,spectrum coupled with texture information classification,selecting the best classification method for near ground hyperspectral image.The results show that the novel spectral index classification method can effectively extract the light wheat and shadow wheat of green stage,jointing stage,booting stage,but in the heading stage,the new spectral index classification method can not extract the light wheat and shadow wheat;texture classification of mathematical morphology can effectively extract the pure the information of wheat green stage,jointing stage,booting stage and heading stage,but could not subdivide pure wheat into the light wheat and shadow wheat;but the new spectral index classification and texture classification based on mathematical morphology are coupled to the near ground hyperspectral image classification can distinguish the wheat at different growth stages the light wheat and shadow wheat,and compare with unsupervised classification and supervised classification,both the overall classification accuracy and Kappa coefficient have greatly increased.The results provide reference using hyperspectral image classification for other crops.Based on the clsssification of sepctra coupled with texture information,we extract the reflectance mean of wheat did not remove background,wheatremove background,light wheat and shadow wheat as four target components’ reflectance.This study selected the target component in visible light and near infrared region and leaf nitrogen content has the highest correlation band building monitoring models predict wheat leaf nitrogen content,which the wheat did not remove background,wheat remove background,light wheat and shadow wheat have highest correlation coefficient bands are 635.06 nm and 938.54 nm,612.20 nm and 939.87 nm,574.29 nm and 917.25 nm,570.25 nm and 970.60 nm.The results show that single-band model based on the wheat remove background,its modeling set and test set in the visible and near-infrared determination coefficient R2 is higher than other target components 0.05-0.25.To further validate the power of each target component predict wheat leaf layer nitrogen conten,the study used normalized difference spectral index(NDSI),ratio spectral index(RSI)and difference spectral index(DSI),for example,to build wheat leaf layer nitrogen content prediction model,the results showed that the spectral indices NDSI,RSI and DSI of wheat remove background forecasting LNC has higher estimation accuracy than other target components,and the predicted results of the 1:1 line slope is close to 1.By using successive projections algorithm(SPA)extracted the characteristics band of the wheat did not remove background,wheat remove background,light wheat and shadow wheat based on wheat leave layer nitrogen content,and obtaining the texture information of each target component characteristics bands based on GLCM,respectively.The spectra information,texture information,texture spectra information of characteristic bands as the independent variable,to build arbitrary two band best vegetation index(NDVI,RVI and DVI).Studies show that when the combination of spectrum and texture information to build spectrum texture index,the accuracy of the each target component to monitor wheat LNC has greatly improved,and the NDSTI based on the wheat remove background monitoring effect is best,its calibration set and validation set of determination coefficient R2 of 0.78 and 0.83,respectively,the relative root mean square error(RRMSE)and relative error(RE)were 0.11 and 0.13,relatively spectral index NDSI,calibration set and test set of R2 were increased by 0.07 and 0.12 respectively,for the texture index NDTI,it increased 0.08 and 0.11.