Inversion Of SPAD And LAI In Cotton Canopy Leaves Based On UAV Multispectral

As an important industrial crop,cotton plays a critically significant role in the development of the world economy and China’s economy.Soil and plant analyzer development(SPAD)and canopy leaves area index(LAI)are important indicators for evaluating cotton growth.Meanwhile,accurate estimation of cotton canopy leaves SPAD values and LAI is a prerequisite for monitoring cotton growth and guiding its production and management.At the moment,remote sensing technology is widely used in crop research,especially UAV remote sensing technology,which is increasingly used for agricultural monitoring due to its advantages of flexibility,ease of operation,high spatio-temporal resolution,etc.Quantitative spectral analysis of cotton chlorophyll and canopy leaves area index using UAV with spectroscopy sensors has become a research hot spot in remote sensing.However,researches on SPAD and LAI UAV near-ground spectral characteristics of cotton canopy leaves in the North China plain area and the best modeling method still require to be strengthened.For the cotton areas of North China Plain in the Yellow River Valley,the article took the cotton field of Dalizhuang,in Xiajin County,Dezhou,Shandong as the test area,the cotton samples were collected from August 3rd to 6th,2019,to measure the value of SPAD,LAI as well as other main agronomic parameters of the cotton,and near-ground multi-spectral images of UAV in the test area was obtained.Then,spectral characteristics of SPAD and LAI were analyzed based on UAV near-ground images to construct spectral indexes and screen spectral parameters.Afterwards,multiple stepwise regression(MSR),support vector machine(SVM)and back propagation neural network(BPNN)were adopted to establish quantitative inversion models of SPAD and LAI,followed by selection of the best model after contrast and comparison.Finally,spatial distribution inversion of SPAD and LAI was conducted in the test area.This study is very helpful for monitoring cotton growth and production in the North China Plain.Specific contents and results are as follows:(1)Spectral characteristics and spectral parameters of SPAD and LAI of cotton canopy leaves in the test area were investigatedThe spectral characteristics of cotton canopy leaves SPAD in the test area was mainly reflected in the red light and red edge bands,of which the red-light band was significantly negatively correlated with cotton canopy leaves SPAD,while the red edge band was positively correlated,and the correlation of the red-light band was higher than that of the red edge band.Based on the correlation analysis,the characteristic spectral parameters of cotton canopy leaves SPAD were selected as(red,r~*reg,(reg-r)/(reg+r),r-g,r/g,(?)).The spectral characteristics of cotton LAI in the test area were mainly in the near infrared(NIR)and red edge bands.The NIR band was significantly positively correlated with cotton LAI and the same with red edge band.However,the sensitivity of the red edge band was lower than that of the NIR band.Based on the correlation analysis,the characteristic spectral parameters of cotton LAI were selected as(nir,r+nir+reg,g~*nir,g+reg+nir,r+nir+g,r+g+nir+reg).(2)SPAD and LAI inversion models of cotton canopy leaves in the test area were establishedBased on the UAV near-ground images,SPAD and LAI inversion models of cotton canopy leaves were established using multiple stepwise regression(MSR),support vector machine(SVM)and BP neural network(BPNN)methods,respectively,with the screened characteristic spectral parameters as independent variables.Comparing the three modeling methods,for SPAD,the BP neural network model was the most accurate and the accuracy of support vector machine model was closer to that of the multiple stepwise regression model.For cotton LAI,the accuracy of support vector machine model was the highest,followed by BP neural network model and multiple stepwise regression model.Through model comparison and analysis,in the inversion of SPAD and LAI of cotton canopy leaves,the accuracy of the two nonlinear models is higher than that of the linear model.Therefore,the optimal inversion model for cotton leaf SPAD was BPNN,and the R~2,RMSE and RPD of the modeling set were 0.747 and 4.568,respectively,and the R~2,RMSE and RPD of the validation set were 0.758,4.142 and 2.135,respectively.The optimal inversion model for LAI is SVM,whose modeling set R~2and RMSE are 0.665 and 4.127 respectively,and the validation set R~2,RMSE and RPD are 0.713,4.653 and 2.575 respectively.(3)The spatial distribution characteristics of SPAD and LAI of cotton canopy leaves in the test area were clarifiedBased on the multi-spectral image of the test area and the corresponding optimal model,regional inversion and hierarchical division of SPAD and LAI of cotton canopy leaves were carried out,and spatial inversion distribution maps of SPAD and LAI of cotton canopy leaves in the test area were obtained.The inversion values and measured values were analyzed mathematically and statistically,and then compared with the spatial interpolation map of measured data.The results showed that the SPAD inversion values of cotton canopy leaves in the test area ranged from 19.3 to 58.6,and concentrated in more than 40.0,accounting for78.72%.In terms of spatial distribution,the SPAD inversion values of cotton canopy leaves were lower in the north and higher in the south.LAI inversion values ranged from 1.12 to8.04 and were concentrated above 4.0,accounting for 61.71%.In terms of spatial distribution,LAI inversion values were also lower in the north and higher in the south.Compared with 94sample points,the overall inversion value is consistent with the measured value,and the proportion of the inversion value and measured value at all levels of sample points is also consistent.The spatial inversion distribution map is highly consistent with the measured values of low in the north and high in the south reflected by the spatial interpolation results of the measured values,which is consistent with the field situation of cotton canopy leaves in the north with poor yellowing growth and cotton canopy leaves in the south with good growth.The inversion results are better.