Research On Maize Growrh Information Monitoring Based On Coupling Feature Spectrometer And Imaging Spectrometer

Maize is one of the three major grain crops in China,its yield and quality directly affects the development of our country’s food security and agricultural production.The content of chlorophyll,plant water content,leaf nitrogen content and other physiological parameters have an important effect on the growth and yield of maize and can provide reference for farmland management,irrigation and fertilization.The hyperspectral technology assessment is to achieve quantitative monitoring of crop physiological parameters with simple and effective technical means for rapid nondestructive.In order to further improve the accuracy of hyperspectral monitoring of physiological parameters of maize,the maize based on field test and indoor test in Northwest China during the different growth period was as the research object.Based on feature spectrometer and imaging spectrometer spectral data,chlorophyll(Soil and plant analyzer development,SPAD)values,plant water content and leaf nitrogen content at different growth periods of maize were measured,the correlation between spectra at different growth stages under different spectrometers and SPAD values,plant water content and leaf nitrogen content were systematic analysed.Hyperspectral estimation models for maize physiological parameters were established and verified based on the characteristic bands,vegetation index and hyperspectral characteristic parameters.The research results can provide theoretical basis and technical support for remote sensing monitoring and quantitative inversion of maize growth in Northwest China.The main conclusions are as follows:(1)At different growth periods,the maize canopy reflectance response characteristics were different.In the "Green Peak" band,along with the development process,the spectral reflectance increased gradually,in the near infrared band,the plants reflectance in tasseling stage and milk stage were larger,in the jointing stage and ripe stage were smaller;from jointing stage to milk stage,leaf SPAD values increased,milk stage to ripe stage,leaf SPAD values decreased,with the growth process increasing,the plant water content showed a gradually decreasing trend.(2)With the correlation analysis between maize SPAD values and canopy SVC(Sp ectra Vista Corporation)spectrum which acquired by feature spectrometer under maize jointing stage,tasseling stage,milk stage and ripe stage showed that the original spectrum and SPAD values respectively in 709 nm,552nm,712 nm and 710 nm reached t he maximum correlation;first derivative spectrum and SPAD values respectively in 752 nm,756nm,760 nm and 749 nm reached the maximum correlation;vegetation indices GRVI,GNDVI,MCARI,TCARI and SPAD values under four maize growth periods had significant correlation;the correlation coefficients between hyperspectral characteris tic parameters λr,Db,SDb,SDg,SDr/SDb,SDr/SDy,(SDr-SDb)/(SDr+SDb)and(SDr-SDy)/(SDr+SDy)and SPAD values reached the absolute value of above 0.7 in the four stag es,thus had great universality.Among the regression models,power function model w hich was built by the first derivative of spectrum at tasseling stage and multiple linear regression model which was built by hyperspectral characteristic parameters at tasselin g stage were the best models of SPAD estimation models.(3)With the correlation analysis between maize plant water content and canopy SVC spectrum,correlation between the original spectrum and plant water content in each growth period were small;first derivative of spectrum and plant water content respectively in 1685 nm,2090nm,2455 nm and 433 nm reached the maximum correlation in each growing period,the maximum correlation coefficients respectively were 0.542,0.570,0.510 and-0.685;new spectral indices which were constructed by correlation matrix,showed that FD2246/2084,FD2234/2028,FD2337/2249 and FD(2341-433)/(2341+433)were the best spectral indices of each growth period,and the correlation coefficients with plant water content were 0.716,0.668,0.726 and-0.888.Among the regression models which were established based on the characteristic bands,water indices and new spectral indices,exponential model which was built by the FD(2341-433)/(2341+433)at the ripe stage and multiple linear regression model which was built by new spectral indices at the ripe stage,were the best models of plant water content estimation models.(4)With the correlation analysis between maize SPAD values and leaf SOC(Surface Optics Corporation)spectrum which acquired by imaging spectrometer,the maximum correlation coefficient between SPAD values and the original spectrum located at 717.17nm(R=-0.567);the maximum correlation coefficient between SPAD values and the first derivative spectrum located at 696.12nm(R=-0.841).Vegetation indice MCARI,showed maximum negative correlation with the SPAD values,the correlation coefficient was-0.830;the hyperspectral characteristic parameters Dy,SDg and(SDr-SDb)/(SDr+SDb)were correlated with SPAD,the correlation coefficients were 0.799,-0.795 and 0.862.Among the regression models which were established based on the characteristic bands,vegetation indices and hyperspectral characteristic parameters,linear model using(SDr-SDb)/(SDr+SDb)as the independent variable and multiple linear regression model using first derivative spectrum as the multivariate were the optimal estimation models.(5)With the correlation analysis between maize leaf nitrogen content and leaf SOC spectrum,SPAD values was most correlated to the original spectrum at 711.90 nm and it was most correlated to the first derivative spectrum at 545.84 nm,the correlation coefficients respectively were-0.530 and-0.667.Vegetation indice GNDVI showed maximum positive correlation with the leaf nitrogen content,the correlation coefficient was 0.608;the hyperspectral characteristic parameters Ro,SDg and SDr/SDb were correlated with leaf nitrogen content,the correlation coefficients were 0.578,-0.635 and 0.717.Among the regression models which were established based on the characteristic bands,vegetation indices and hyperspectral characteristic parameters,linear model using SDr/SDb as the independent variable and multiple linear regression model using hyperspectral characteristic parameters as the multivariate were the optimal estimation models.