Study On Estimation Model Of Water And Nitrogen Content In Wheat Flag Leaf Based On Digital Image Processing

Wheat is one of the most important crops in the world and an important food crop in China,which is grown in many provinces,cities and autonomous regions across China.People over-apply fertilizer and irrigation to increase wheat production,causing waste of resources and polluting the environment simultaneously.Accurately obtaining the water and nitrogen content of wheat and developing a scientific water and fertilizer transportation plan is the key to scientific management and efficient production.Flag leaf is the most important blade of wheat,which has great influence on the growth and development of wheat,in the middle and late period of wheat growth,the water and nitrogen content of flag leaf directly affects the quality and yield of wheat.This study analyzed the relationship between water and nitrogen content of wheat flag leaf under drip irrigation and color characteristic parameters of the digital image in southern Xinjiang,and established an estimation model of water and nitrogen content of wheat flag leaf based on color characteristic parameters of the digital image.It realizes the accurate and efficient estimation of water and nitrogen content of wheat flag leaf,and provides the theoretical basis for accurate water and fertilization cultivation of wheat.The main results of the research are as follows:（1）Screening of color characteristics parameters with high sensitivity to water and nitrogen content of wheat flag leaves was completed.21 feature parameters were selected from the perspective of image color features,and Pearson correlation coefficient method and Principal component analysis method were used to screen the selected parameters.By Pearson correlation coefficient method,12 color parameters with high correlation with nitrogen content of flag leaf,such as（r-g-b）/（r+g）,and 7 color parameters with high correlation with the relative water content of flag leaf,such as red green reflectance index（RGRI）,were screened;by principal component analysis,4 color parameters,such as R,G,B and super red index（E×R）,were screened.（2）The best estimation model for the nitrogen content of wheat flag leaf was constructed.Four modeling methods,linear regression,partial least squares regression,support vector machine regression,and principal component regression,were used.The models were constructed with the color characteristic parameters obtained by the two screening methods as independent variables,respectively,and the models were evaluated with the coefficient of determination（R²）,root mean square error（RMSE）and relative error（RE）as criteria.It was shown that the support vector machine regression model constructed based on the color characteristic parameters screened by the Pearson correlation coefficient method had the best ability to estimate the nitrogen content of wheat flag leaves.The R² RMSE,and RE of the model were 0.914,1.584 g/kg,and 12.876%,respectively.（3）The best model for estimating the relative water content of wheat flag leaves was constructed.The model was constructed and evaluated in the same way as the best estimation model of nitrogen content of wheat flag leaf.It was found that the linear regression model constructed based on the color characteristic parameters screened by Pearson correlation coefficient method had the best ability to estimate the relative water content of wheat flag leaves.The model was Y=2 613.769+0.084×x₁-2 383.805×x₄-199.786×x₁₂-270.221×x₁₄-2 511.235×x₁₇+690.376×x₁₉+664.235×x₂₀。The R² RMSE,and RE of the model were 0.791,2.498%,and 2.221%,respectively.（4）Through independent sample test,the estimation models were verified,and it was found that each estimation model had high stability.Among them,the best estimation model of nitrogen content of wheat flag leaves verified R² was 0.889,RMSE was 1.605 g/kg,and RE was 15.362%;the best estimation model of relative water content of wheat flag leaves verified R² was 0.742,RMSE was 2.857%and RE was 2.592%.It indicated that the water and nitrogen content of the flag leaf of wheat during growth can be effectively and accurately estimated by the above two models.