Research On Estimation Model Of Corn Leaf Area Index Based On UAV Image

Leaf area index（LAI）is an important index of crop growth and yield.Because of its non-destructive and high-throughput characteristics,remote sensing technology is called the only scientific method to carry out rapid,efficient,economic and nondestructive LAI detection research.At present,LAI estimation based on traditional remote sensing technology has been widely used in precision agriculture and achieved results.Because of the long revisit period,the mutual restriction of image spatial resolution and spectral resolution,cost-effectiveness and other factors,the ground-based survey method and satellite remote sensing method are difficult to meet the needs of precise real-time LAI estimation.In recent years,the deep cross fusion of UAV technology and hyperspectral image has become one of the main ways to obtain the hyperspectral remote sensing data of centimeter level super-high resolution and hour level real-time response of crops in large area at present and in the future,which provides a new data support for real-time and accurate estimation of crop LAI.Machine learning method has better adaptability to nonlinear problems and can make better use of spectral information,so it is introduced into the study of LAI estimation.The method of UAV hyperspectral data integration machine learning is used to estimate the LAI of crops,so as to better achieve the purpose of crop disaster prevention and refined field management.Maize is the most widely planted food crop in China,and its physiological characteristics have strong representativeness.The estimation of maize LAI has strong migration and reference significance for other food crops.In this study,corn field cultivation experiments were carried out in Yucheng City,Shandong Province.This experiment obtains the corn hyperspectral remote sensing image data and simultaneously measures the land-based LAI data.The problem of LAI estimation of Maize Based on hyperspectral data of UAV is systematically explored.The main research contents and corresponding conclusions are as follows:1.For the acquisition of LAI response band,the PROSAIL model was used to simulate the canopy reflection of maize.Combined with correlation analysis,the characteristic response bands of maize LAI were 516,636,702,760 and 867nm.The correlation between the characteristic response band and maize LAI was-0.82^**,-0.84^**,-0.84^**,0.66^**,and 0.69^**,respectively（P<0.01）.Among them,702nm and636nm have the largest correlation coefficient with maize LAI,which has the potential to improve the accuracy of LAI estimation.2.Comparing the four methods of hyperspectral data change,corn hyperspectral location and area feature information,PROSAIL model response band and vegetation index,it is found that the combination of two or more bands can enhance the response to maize LAI,including NDVI,RVI,r NDVI,m NDVI,EVI2 and OSAVI.The vegetation index is highly correlated with maize LAI（R²=0.85-0.88）,which is suitable for maize LAI estimation.3.The prediction ability of LAI estimation model based on machine learning algorithm is obviously better than that of empirical statistical method,and the prediction accuracy of LAI estimation model based on RFR（C-R²=0.93,V-R²=0.91,C-RMSE=0.06,V-RMSE=0.07）is better than that of other machine learning models,which is most suitable for the inversion of LAI of crops in typical agricultural areas of Huang Huai Hai Plain.