Research On Wheat Yield Estimation Model Based On UAV Image And Physiological Index

In order to grasp the dynamic information of wheat production in a timely and accurate manner,this study is based on the UAV equipped with a high-definition digital camera as the remote sensing platform,and the Sanping Experimental Base of Xinjiang Agricultural University is selected as the research area to obtain the UAV visible light image splicing generated during the key growth period of wheat.The high-definition digital orthophoto map(DOM)of the growth period analyzes the correlation of 4 indicators(leaf area index,leaf nitrogen content,leaf water content and chlorophyll)and 10 vegetation indices with wheat yield respectively.The physiological index and vegetation index that are most sensitive to yield were screened out,and the applicability of three modeling methods(univariate linear regression UR,multiple stepwise regression SMLR and principal component regression PCA)in estimating the yield of wheat during each growth period was compared,and the wheat yield was obtained.Optimal yield estimation model for optimal fertility period.conclusion as below:(1)The two types of variables of vegetation index and physiological index in each growth period of wheat are basically consistent with the change characteristics of the correlation between wheat yield,and the correlation shows the characteristics of flowering period> heading period> filling period> 7 days after filling period> maturity period;The change trend of the correlation between the physiological index and the vegetation index in each growth period and the yield of wheat is different;the correlation between the physiological index and the yield is shown as the trend of leaf nitrogen content> LAI> SPAD> leaf water content,while the vegetation index is in The performance is different in each period.(2)Through the significance test,it can be obtained that in each growth period of wheat,except for the poor correlation between the single-band vegetation index and the yield,the rest of the vegetation index and the yield all show a strong and extremely significant relationship.Among them,G,Ex G and Ex R are related to the yield per unit area.There is a negative correlation between them,and other vegetation indices are all positive correlations.Compared with other growth periods,the correlation between flowering and yield is the highest,and the maturity is the lowest.(3)Among the wheat yield estimation models for each period constructed based on three variables,the flowering period model constructed by combining physiological indicators and vegetation index as independent variables has the highest yield estimation accuracy,and its R2,RMSE and n RMSE are 0.906,212.45 kg/hm2 and 212.45 kg/hm2,respectively.8.02%;Compared with the R2 of the optimal yield estimation model constructed with the vegetation index as the independent variable,the R2 increased by22.85%,28.50%,and 35.6%;the accuracy of the model constructed with the physiological index as the independent variable increased by 2.05% respectively.27.76% and 24.46%.(4)Three modeling methods are used to explore the applicability of wheat yield estimation in various periods.The results show that,compared with multiple stepwise regression and principal component regression,the structure of the unary linear regression model is simpler and fails to fully summarize the relationship between the variables and the yield.For the quantitative relationship,except for the PCA model,which is the optimal yield estimation model for 7 days after the filling period,the SMLR model has the best yield estimation effect for the rest of the growth period.(5)The remaining sample data is used to verify and evaluate the accuracy of the yield estimation model for each growth period.The results show that the 1:1 scatter plot of the actual production value and the predicted value is basically consistent in each growth period,and the yield estimation accuracy of each growth period model is present.Flowering period>heading period>filling period>7 days after filling period>mature period.R2 is the highest in flowering period,and R2 is the lowest in mature period.