Study For UAV-based Remote Sensing Monitoring Main Growth Parameters Of Winter Wheat

Wheat is one of the three maj or cereal crops in the world and a key component in ensuring food security.The growth parameters of wheat are important indicators for monitoring its health status.Using unmanned aerial vehicle(UAV)remote sensing technology to monitor the growth parameters of wheat over large areas in real-time,quickly and non-destructively is beneficial for increasing wheat yield and ensuring food security.Traditional manual surveys are time-consuming,labor-intensive,and costly,and are difficult to meet the requirements of precise management.UAV remote sensing technology can monitor the growth parameters of wheat during critical growth stages,providing technical support for precise management and decision-making in the field.However,at present,the accuracy of UAV remote sensing for monitoring multiple growth parameters of wheat still cannot meet the needs of actual production.Previous studies have extensively investigated the monitoring of growth parameters during single growth stages,but research on the accuracy comparison and algorithm evaluation of UAV remote sensing for monitoring multiple growth parameters of wheat at different growth stages is relatively scarce.There is even less research on the practicality and predictability of UAV remote sensing monitoring models constructed by various algorithms through systematic analysis and evaluation.To improve the accuracy of UAV remote sensing monitoring of wheat growth parameters and provide decision support for wheat production and precision management,this study focuses on the SPAD(Soil Plant Analysis Development)value of leaves,leaf nitrogen content(LNC),and above-ground biomass(AGB)growth parameters during the jointing stage,booting stage,heading stage and flowering stage in wheat fields in the Yizheng area of Yangzhou City,Jiangsu Province,China,from 2020 to 2021.Based on the multispectral image information of UAVs during each growth period,the feasibility and predictability of singlefactor and multi-factor models are compared to improve the accuracy of monitoring and prediction of wheat growth parameters.The aim of this study was to improve the precision of UAV remote sensing monitoring and forecasting of main growth parameters of wheat in multiple growth periods at field.The main contents and results of this study are as follows:(1)Multiple regression algorithms were used to construct the UAV remote sensing monitoring model of wheat main growth parameters in the main growth period.The growth parameters of wheat leaf SPAD,LNC and AGB in the main growth periods(joining,booting,heading and flowering)were taken as the research objects.50%of wheat data from 2020-2021 was taken as the training data,the rest was taken as the test data,the growth parameters were taken as the output variables,and the feature parameters extracted from multispectral images were taken as the input variables.Three algorithms,linear Regression(LR),PLS(Partial Least Squares)and RF(Random Forest),were used to establish UAV remote sensing estimation models for main growth parameters of major growth periods.Combined with the two evaluation indexes,the determination coefficient R2 and the root mean square error RMSE,and the 1:1 relationship diagram between the measured value and the predicted value of the model,the practicability and predictability of the model constructed in this study were systematically evaluated and compared,and then the best model was determined for each growth period.(2)The results showed that LR model had the best performance at jointing stage(R2=0.23,RMSE=3.80)and heading stage(R2=0.64,RMSE=5.12)for SPAD value of wheat leaves.PLS model showed the best performance at booting stage(R2=0.51,RMSE=4.01).RF model had the best performance at the flowering stage(R2=0.78,RMSE=1.92).For wheat LNC,RF model performed best at jointing stage(R2=0.67,RMSE=7.35 g/kg),booting stage(R2=0.43,RMSE=8.04 g/kg)and heading stage(R2=0.70,RMSE=4.16 g/kg),while PLS model performed best at flowering stage(R2=0.67,RMSE=5.67 g/kg).For wheat AGB,RF model performed best at jointing stage(R2=0.89,RMSE=241.94 kg/ha),booting stage(R2=0.87,RMSE=353.33 kg/ha),heading stage(R2=0.74,RMSE=958.42 kg/ha)and flowering stage(R2=0.84,RMSE=607.06 kg/ha).Overall,compared with other methods in this paper,the model built based on RF algorithm can better monitor the leaf SPAD value and AGB of the main growth stages of wheat at the joining stage,booting stage,heading stage and flowering stage by UAV remote sensing.The model based on LR can better monitor the SPAD value of wheat at jointing stage and heading stage by UAV remote sensing.The model based on PLS algorithm can better monitor the SPAD value of wheat at booting stage and LNC at flowering stage.