| With the widespread popularization of UAV remote sensing technology,how to accurately grasp crop growth and phenotypic parameters has gradually become the key to fine agricultural management.In order to efficiently and accurately estimate the height(H)and height of crops at different growth stages,Leaf area index(LAI)phenotypic traits.In this study,wheat H and LAI were used as research objects,and the acquired UAV images were combined with Ground Control Points(GCP)to generate high-resolution Digital Orthophoto Maps(DOM)and digital surfaces for wheat growth stages Digital Surface Model(DSM),and through three regression methods and comprehensive evaluation indicators(R2,RMSE,and nRMSE)to explore the best H and LAI estimation methods in different periods.The main research conclusions of this paper are as follows:(1)The plant height data of wheat at different growth stages extracted based on DSM are smaller than the measured values,but the overall deviation is not large.The fitting accuracy of the measured and predicted values of H is good in different growth stages,and the precision in the whole growth stage is the highest(modeling set R2=0.85,validation set R2=0.87).In different periods,the accuracy during the grouting period is highest(modeling set R2=0.72,validation set R2=0.79),and the maturity period is lowest(modeling set R2=0.68,validation set R2=0.69).(2)In the study of LAI estimation model based on univariate linear regression,the measured LAI values in each growth period are consistent with the growth and change law of wheat LAI,and the vegetation index and LAI in each period have strong significance.Except for the maturity period,the absolute values of the correlation coefficients of ExGR,ExR,MGRVI,NGRDI,RGRI and LAI all reached above 0.7(P<0.001).Among them,the overall significance was the strongest at the grouting stage,and the overall significance was weaker at the mature stage.In the modeling set,the VARI fit was highest during the grouting stage(R2=0.64),and the RGBVI fitting was the lowest(R2=0.28)in the mature stage.In the validation set,the VARI fit was highest in the jointing stage(R2=0.68),and the mature stage RGB VI has the lowest fit(R2=0.45).(3)In the study of the LAI estimation model based on multiple stepwise regression,as the number of input independent variables of each growth period estimation model gradually decreases,the significance of each vegetation index is significantly enhanced,and the complexity of the model is getting lower and lower.In the modeling set,the fitting degree at the jointing stage was the highest(R2=0.67)and the maturity stage was the lowest(R2=0.43).In the validation set,the fitting degree at the jointing and grouting stages was the highest(R2=0.68),and the maturity stage was the lowest(R2=0.53).(4)In the study of the LAI estimation model based on partial least squares regression,as the number of input principal components in each growth period estimation model gradually decreases,the complexity of the model decreases.In the modeling set,the fitting degree at the jointing stage was the highest(R2=0.66)and the maturity stage was the lowest(R2=0.45).In the validation set,the fitting degree at the jointing stage was the highest(R2=0.68)and the maturity stage was the lowest(R2=0.53).(5)Based on the three LAI estimation models constructed,it is concluded that the three estimation models in each growth period have achieved high fitting accuracy and stability,of which the jointing and grouting periods have the highest accuracy(R2 is above 0.60).Analysis of 15 validation sets found that the fitting accuracy of the measured and predicted LAI in each growth period was high,and both had high accuracy and stability.Among them,the multivariate stepwise model R2 at the jointing and filling stages was 0.68,and the maturity univariate linear model has a minimum R2 of 0.52.According to the evaluation index,multiple stepwise regression is the optimal LAI estimation model for each growth period of wheat.According to the DOM observations of the wheat in different growth stages generated by splicing,the overall growth of wheat in the filling stage is the best and the vegetation coverage is the highest.Layer coverage is reduced. |
