Study On Hyperspectral Model Of Physiological And Biochemical Parameters Of Winter Wheat At Jointing Stage Under The Influence Of Frost

Frost is one of the main agrometeorological disasters in winter wheat production.In order to improve the monitoring accuracy and accuracy of winter wheat under the influence of frost,this study investigated the use of hyperspectral remote sensing to monitor plant physiological and biochemical parameters.The experiments were performed on winter wheat at the jointing stage as the research object and carried out in the agricultural extraction garden and meteorological laboratory of Anhui Agricultural University in 2021.Five wheat varieties conventionally planted in Jianghuai area were selected for the experiment:Xunong029(A1),Weilong 169(A2),Wanken 869(A3),Hengjinmai 8(A4)and Xumai 35(A5).Three stress temperature treatments were also conducted-1.0°C(L1),-3.0°C(L2),-5.0°C(L3).The study methodology involved hyperspectral data acquisition,measuring chlorophyll relative content(SPAD)and soluble sugar content,determining physiological and biochemical indexes such as peroxidase(POD),analyzing the obtained hyperspectral data and physiological and biochemical indexes of the winter wheat leaves,studying the change trend of physiological and biochemical indexes of winter wheat before and after frost.After the original spectra were pretreated with first derivative(FD),Hilbert transform(HT),Savitzky Golay filter(SG)and fast Fourier transform(FFT),characteristic spectrum information was extracted by statistical analysis methods such as kernel principal component analysis(KPCA),competitive adaptive reweighting algorithm(CARS),successive projection algorithm(SPA),and using machine learning algorithms such as LIBSVM,random forest(RF),extreme learning machine(ELM)and back propagation neural network(BPNN)to build estimation and hierarchical monitoring models based on the physiological and biochemical indexes and hyperspectral data.The main conclusions of this study are as follows:(1)With the decrease of temperature,the SPAD value of winter wheat varieties mainly increased first and then decreased,and the soluble sugar content and POD activity mainly increased.The maximum soluble sugar content of A2,A3,A4 and A5 varieties and the maximum POD activity of A2,A3 and A4 varieties appeared in L3 treatment.There were significant differences in physiological and biochemical indexes among varieties under low temperature stress.There was a certain correlation between SPAD value and soluble sugar content,and the correlation coefficient was 0.52.(2)The shape and change trend of the spectral curve of the winter wheat leaves before and after frost were basically similar.With the change of low temperature level,the higher the SPAD value,the smaller the reflectivity in visible band and the larger the reflectivity in near-infrared band;The lower the SPAD value,the greater the reflectivity in visible band and the smaller the reflectivity in near-infrared band.The soluble sugar content and POD activity of winter wheat in A3 and A4 groups increased significantly under different low temperature levels,and the SPAD value increased compared with that before frost.The spectra after FD and HT treatment changed greatly compared with the original spectrum,highlighting the band characteristics and improving the correlation between physiological and biochemical indexes and spectra.(3)The original spectrum of winter wheat leaves was pretreated to establish correlation with physiological and biochemical indexes,and the spectral characteristic bands with high correlation were extracted.After dimensionality reduction of characteristic bands,LIBSVM,RF,ELM and BPNN models were established and verified respectively.The results showed that the effect of the winter wheat SPAD monitoring model constructed by FD-CARS-RF combination was best,with a prediction set R~2of 0.721 and RMSE of 4.259.The effect of the winter wheat soluble sugar monitoring model constructed by FD-CARS-RF combination was similarly the best,with a prediction set R~2 of 0.546 and RMSE of 1.642.The effect of the winter wheat pod monitoring model constructed by HT-KPCA-RF combination performed best,with a prediction set R~2 is 0.256 and RMSE of 1.740×10~4.(4)After preprocessing and dimensionality reduction of the spectral data of winter wheat leaves,the characteristic bands were extracted.The low temperature levels of winter wheat was classified and evaluated by LIBSVM,RF,ELM and BPNN models respectively.The results showed that HT-SPA-LIBSVM,HT-SPA-BP and FD-SPA-ELM were better among the winter wheat frost classification detection model combinations.Their accuracies on the test set were 0.9050,0.8825 and 0.8800,respectively,and their standard deviations were0.0510,0.0487 and 0.0367,respectively.