Research On Maize Lai Monitoring Method On UAV Multispectral Remote Sensing

As one of the representatives of crops,corn is second only to wheat and rice.In China,corn is not only an important food crop and industrial raw material,but also an important part of China's national economy.Therefore,monitoring the growth dynamics of corn is the focus of China's maize crop research at this stage.Leaf area index(LAI)is a parameter that reflects the number of crops and effective monitoring of pests and diseases.It is an important indicator to describe the growth status of vegetation,and it is of great significance in predicting crop growth trends and yield.In this paper,we use the corn of the Zhaojunzhen Precision Irrigation Experiment Base in Dalateqi,Inner Mongolia as the research object in 2017,and use the independently developed six-rotor UAV equipped with the RedEdge five-band multispectral camera to obtain remote sensing images of the experimental area and use the LAI-2200 crown.Measured leaf surface area index of ground corn by a layer analyzer,based on five commonly used vegetation indices(normalized difference vegetation index NDVI,optimized soil regulated vegetation index OSAVI,soil conditioned vegetation index SAVI,enhanced vegetation index EVI,renormalized The vegetation index(RDVI)was used to invert the leaf area index of field corn under different water stress treatment levels.The main research content of this article is as follows:(1)For the problem that the remote sensing image is distorted due to different flight parameters of the UAV,the best flight parameters are obtained by comparing multiple experiments in the process of acquiring corn remote sensing image data.The results show that when the UAV flight altitude is 70 m,the heading overlap is 80% and the sideward overlap is 70%,the best remote sensing image data can be obtained.(2)Using the linear regression method,a regression model of vegetation index and leaf area index were constructed under water stress and no water stress conditions respectively.The accuracy was evaluated by using the determination coefficient and root mean square error to find a better index of leaf area index modeling.The results showed that under the condition of no water stress,the best models in each stage were the SAVI-LAI model at the jointing stage,the EVI-LAI model at the growth stage and the EVI-LAI model at the mature stage with good precision,and the determination coefficients were 0.646,0.736,and 0.771.Under the condition of water stress,the best model in each stage is: the jointing stage NDVI-LAI model,the growth period EVI-LAI model and the mature period EVI-LAI model have good accuracy,the determination coefficients are 0.863,0.751 and 0.645.(3)Using multivariate linear regression,a regression model was established for the observation of multiple vegetation indices and leaf area indices under different water stress conditions.It was concluded that under the condition of no water stress: the regression coefficient of the regression model at the jointing stage is 0.819,the regression coefficient of the growth stage is 0.832,and the maturity coefficient of the regression model is 0.834.Under conditions of water stress: the coefficient of regression of the regression model at the jointing stage is 0.926,the coefficient of the regression model of the growth stage is 0.849,and the coefficient of regression of the regression model at maturity is 0.770.(4)Using support vector machines,the combination of EVI,OSAVI,SAVI,NDVI,and RDVI vegetation indices was used as the input,and the leaf area index was the output,and a support vector machine model was constructed.The model obtained from the training was used to predict and found that the support vector machine model was effective.Under the condition of no water stress,the coefficient of determination reached 0.853.Under the condition of water stress,the coefficient of determination reached 0.851.(5)The obtained linear regression model,multiple linear regression model and support vector machine regression model were used to evaluate the accuracy.The SVM model was found to be the best inversion model.The accuracy reached 91.3% under the conditions of no water stress,but the accuracy reached 90.5% under the condition of water stress,which was higher than other models..Therefore,using the support vector machine model to perform LAI inversion on corn remote sensing images during the observation period,the LAI spatial distribution map was obtained.According to the study,it is feasible to use UAV multi-spectral remote sensing images to rapidly and widely monitor corn field under different water stress conditions.This provides a basis for guiding crop irrigation water use and effectively promotes the research of crop remote sensing theory in China application,after comparison,it was found that the leaf area index of maize under water stress varied from 0.9 to 1.7.