Hyperspectral Characteristics Of Winter Wheat And Remote Sensing Monitoring Of Its Agronomic Parameters

The method of hyperspectral estimation of winter wheat leaf chlorophyll,anthocyanin and leaf area index was discussed to provide a theoretical basis for the efficient,non-destructive and large-area monitoring of winter wheat chlorophyll,anthocyanin and leaf area index.In this paper,winter wheat in Guanzhong area of Shaanxi Province is taken as the research object.SVC hyperspectral data of jointing,heading,flowering and filling stages,drone hyperspectral data of heading and flowering,and chlorophyll and flowers at each growth stage are measured Values of green pigment and leaf area index,analysis of leaf wheat spectral changes in different growth stages,and their correlation with SPAD,Anth,and LAI,and construct a univariate estimation model based on sensitive bands,new spectral indices,and a multivariate estimation model based on PLS,and Verify the accuracy of the model;use the hyperspectral data of the drone to make inversion maps on the drone images,and perform a secondary accuracy test with the samples that did not participate in the modeling.The main findings are as follows:(1)Hyperspectral characteristics of winter wheat canopy leaves:the spectral"reflection peak"of each growth period appears near 550 nm,the peak is about 0.16-0.2,and the reflectance increases rapidly at the band of 680-1000 nm,and finally at the band of 720-1000nm Form a highly reflective platform with a reflectivity greater than 0.4.In the first derivative spectrum,the red edge band is the maximum.UAV hyperspectral characteristics of winter wheat leaves:visible light range,spectral"reflection peaks"at various growth stages appear at 550 nm,and"absorption valleys"at around 670 nm;reflectance increases rapidly at 680-750 nm band range,at 800 nm The band reaches its highest near the band,and then begins to decrease.(2)Compared with the original spectral model,the model established by the first-order derivative spectrum sensitive band has greatly improved the quasi-precision of prediction,and the flowering period model has the best prediction effect.In the SPAD model,the R~2of the original spectrum and the first derivative spectrum model at flowering stage is 0.7295,0.8961,and the RMSE is 2.9425,1.8607;in the Anth model,the R~2of the original spectrum and first derivative spectrum model at flowering stage is 0.7562,0.8524,the minimum RMSE is 0.0082,0.0064.In the LAI model,the R~2and RMSE of the original spectral model at flowering stage were 0.4201 and 0.5982,respectively,and the R~2and RMSE of the first derivative spectrum were 0.5753 and 0.5695.It shows that the measurement of chlorophyll,anthocyanin and leaf area index has a high accuracy at the flowering stage.When the accuracy permits,the prediction model constructed by the first derivative sensitive band at flowering stage can be used as a monitoring parameter for winter wheat a quick way.(3)The correlation between winter wheat chlorophyll,anthocyanin content,leaf area index and new spectral index is significantly higher than that with traditional spectral index.In the chlorophyll unary prediction model constructed with the new spectral index,the DSI model at flowering stage has a higher R~2of 0.8885 and the smallest RMSE is 1.8782;in the anthocyanin model,the RSI model at flowering stage has the highest decision coefficient of0.8679,The RMSE is the smallest,0.0060;in the LAI model,the R~2model at flowering stage R~2is 0.6586,and the RMSE is 0.4388.It shows that the new spectral index univariate prediction model of each growth period has high prediction accuracy,and can be used as a good model for determining the agronomic parameters of winter wheat.(4)SPAD,Anth and LAI models based on PLS in each growth period have the highest R2 and the best prediction effect.In the SPAD and Anth prediction models,the R~2at the jointing stage is 0.1051,0.1627,the accuracy is very low,but the PLS model R~2reaches0.6240 and 0.7010,the decision coefficient of the prediction model at the heading stage is increased from 0.3759,0.3207 to 0.8275,0.8804,and flowering The R~2of the PLS model in the period and the filling period is also very high;in the LAI model,the R~2of each growth period is increased from 0.1199,0.3679,0.4201 and 0.3760 of the original spectrum to0.6411,0.7179,0.7311 and 0.6672 of the PLS model.Therefore,the multivariate prediction model based on PLS is the optimal model for monitoring various agronomic parameters.(5)Among the models built with hyperspectral UAVs,the R~2of the first-order derivative spectrum prediction model is significantly better than the original spectrum,and the prediction effects of the SPAD,Anth,and LAI models are all better at the flowering stage;the agronomic parameters of each growth stage Compared with the new spectral index,the correlation with the new spectral index is greatly improved;compared with the index model,the PLS model has an increased R~2and a lower RMSE,indicating that the quasi-accuracy of model monitoring is improved.In the chlorophyll prediction model,the highest R~2at flowering stage is 0.6613,and the RMSE is 2.1381;the anthocyanidin prediction model is better at the flowering stage model,R~2and RMSE are 0.6161,00047;the PLS model at the flowering stage of the leaf area index is 0.6431,RMSE is 0.3028,both with high R~2and low RMSE.It shows that the flowering stage is the optimal growth stage for monitoring plant agronomic parameters.(6)In the UAV hyperspectral image inversion mapping,the average value of the predicted value of the PLS model for each agronomic parameter model in each growth period is closest to the actual measured value;when using the verification set 2 data to verify the accuracy of the mapping result In each model,the PLS model also has the best fitting accuracy and the best prediction effect,which can accurately reflect the SPAD,Anth and LAI spatial distribution of winter wheat in the study area.