Research On SPAD Value And LAI Inversion Of Cotton Leaves Based On Drone Multispectral Images

This article explores the cotton fields at the Agricultural Teaching and Research Practice Base of Tarim University in Kuerle District,Xinjiang Uygur Autonomous Region.Three different unmanned aerial vehicles(UAVs)with sampling flight heights of 30 m,60m,and 100 m were used to collect UAV multispectral images of the cotton fields.The relative chlorophyll content(SPAD)and leaf area index(LAI)data of the ground cotton fields were also sampled.Pre-processing was carried out on the UAV visible light images,and 22 visible light spectral index parameters were extracted.Pearson correlation analysis was used to screen the spectral parameters for inversion model building.The support vector regression(SVR)model,random forest regression(RFR)model,and XGBoost regression(XGBR)model were used to construct the cotton SPAD and LAI inversion model.The specific research results are as follows:(1)Through the analysis of multispectral original bands,it was found that the near-infrared band and the green band are most sensitive to cotton SPAD value,and RESR(red edge ratio vegetation index)showed the best correlation with cotton SPAD at a flight height of 60 m,with a correlation coefficient is-0.814.Similarly,in the correlation analysis of cotton LAI,it was found that visible light containing near-infrared bands in multispectral original bands shows a more sensitive correlation relationship with LAI,and RVI(ratio vegetation index)showed the best correlation with cotton LAI at a flight height of 100 m,with a positive correlation of 0.851.(2)The spectral parameter combinations selected based on correlation analysis were modeled using support vector machine regression model,random forest regression model,and XGBoost regression model.The optimal unmanned aerial vehicle(UAV)flight height for cotton SPAD inversion research was determined to be 60 m,with the optimal model algorithm being the random forest regression model and the optimal spectral parameter combination being RESR and RENDVI.The optimal UAV flight height for cotton LAI inversion research was determined to be 100 m,with the optimal model algorithm being the XGBoost regression model.(3)By using the best model combination for cotton SPAD and LAI inversion,spatial distribution inversion maps of cotton fields in the experimental area were obtained.It was found that in the SPAD inversion map results,the cotton SPAD values in the research area were mainly concentrated between 7.08 and 88.58,which is consistent with the actual sampling data.In the LAI inversion map results,the cotton LAI values in the research area were mainly concentrated between 6.38 and 21.73,which is consistent with the actual sampling data.The cotton leaf SPAD model based on the random forest regression model and the cotton LAI model based on the XGBoost regression model have good inversion effects.