Hyperspectral Response Mechanism Analysis Of Maize Population Growth Index Under Lodging Stress

Maize lodging is caused by the external factors of maize stalks from the natural upright state to the permanent dislocation of the crop affected by the phenomenon which is a common problem of the maize yield.As the largest food crops of China,mazie lodging gradually become one of main factors which limiting the high and stable yield.Hyperspectral remote sensing technology is an important direction of quantitative remote sensing research in modern agriculture.It has its own characteristics such as high resolution,strong continuity and large amount of information.It can carry out on the disaster and growth monitoring,yield impairment assessment with real-time,fast and non-destructive.Therefore,it has an important significance of accurate,fast and effective monitoring of maize lodging in disaster monitoring,yield impairment assessment and so on through hyperspectral remote sensing technology.In this study,the district control lodging experiment was carried out from 2015-2016 in the Xiaotangshan National Precision Agriculture Demonstration Research Base,and simulated two growth period(tasseling period,filling period)of the three types of lodging situation(stem lodging,stalk lodging,root lodging).Analyzing hyperspectral variation of different lodging states in two growing stages and constructing the Leaf area density(LAD)and canopy chlorophyll density(CCD)were,the results were analyzed based on the mathematical statistics analysis method.And gradually realized the hyperspectral inversion of the lodging indicators of mazie based on mathematical statistics analysis algorithm,the main contents and conclusions are as follows:(1)From the comparative analysis of the original canopy spectrum,the differential transformation spectrum and the spectral characteristic parameters with the maize in different growth stages and different lodging treatments,it was found that all the lodged maize was increased compared with the non-lodged maize in the visible light and near infrared bands.The more serious the lodging,the greater the increase,three types of lodging treatment increased the maximum for the root lodging treatment,followed by stalk lodging treatment,stem lodging treatment the smallest;The increase in the visible band of the spectral reflectance of the three types of lodging treatment is greater than that of the near infrared band.With the recovery of the lodging maize,the increase in the rate of visible and near infrared bands is gradually reduced;There has been a "blue shift" phenomenon in three types of lodging treatment in two grown stage,and the more serious lodging,"blue shift" phenomenon more obvious,"red edge" amplitude and "red edge" area larger.With the late recovery of three types of lodging treatment,the phenomenon of "blue shift",the "red edge" amplitude and the "red edge" area gradually decrease.The latter part of the ability among which the strongest recovery ability is stem treatment,followed by root,The worst for stem treatment;The strongest ability of late recovery is the stem lodging treatment,followed by the root lodging treatment,the worst for the stalk treatment;(2)To construct a disaster indicator that can characterize the change of population structure of lodging maize and the disaster index of the physiological change of the maize population,Sensitive band screening and modeling through the correlation analysis and the best exponential method of the original spectrum,the first order differential spectrum and the two indexes.Ranking the lodging disaster index which used to inverse the spectral index through the GRA and correction method,combined with AIC for partial least squares modeling.Based on enumerated method and adjR2 index,the best modeling index is selected and the model construction of disaster index is realized by LS-SVM and RF algorithm.All models were verified by model evaluation indicator of modeling R2 and verify RMSE,found that the optimal model which based on the machine learning algorithm,modeling R2 reached 0.9 or more,verify RMSE is about 1 ~ 1.5.In this study,analyzed the spectral characteristics of different lodging types in different periods based on residential control simulation experiment,constructed the lodging disaster index and constructed the remote sensing inversion model of lodging disaster index based on the mathematical statistics algorithm,which is useful for the application of hyperspectral techniques in the monitoring of maize lodging,provide the theoretical basis for the lodging post-disaster assessment and production impairment.