Study On Estimation Of Maize Leaf Area Index Models Based On Hyperion Data

Maize is one of the representatives of crop, and the area ranks the third in cropplanting areas, which is less than the wheat and rice area. Corn is not only animportant food crops but also an important feed and industrial raw materials. its valueis an important part of national economy. Therefore, research on the maize growth isvery important. The Leaf Area Index (LAI) which is not only the efficient parameterof estimating maize biomass and monitoring maize plant diseases and insect pests butalso one of important indicators of the Growth status and the output forecast. So it hasan important effect on monitoring maize quantitatively using remote sensing.The experiment area locates in Hexin and Helong town of the western ofChangchun City in Jilin province and30samples are selected to measure maize LAIon the ground using the WinScanopy2010a for Canopy analysis instrument. GPSmeasurement techniques is used to correct the Hyperion image and extract Spectralindex from the image. The correlation between LAI values and index is analyzed.Spectral indicators are selected by using MATLAB software considering the featuresnumerious bands, high spectral resolution and narrow band width. A statisticalregressing model and a multiple stepwise regression model based on the principalcomponent method are constructed using the spectral indicators as independentvariables. The test samples are used to validate and evaluate the accuracy of themodels, and the best estimation model is selected. The LAI is inverted based on thebest model. The experiment confirmed that Hyperspectral remote sensing imagescould express the spatial distribution and growth of maize by estimating andmonitoring LAI accurately and quickly within a large area. The results show that:The relationship between canopy reflectance and LAI is analyzed with ten singlevariant and multivariant regression models, and the derivatives of reflectancespectrum can enhance the correlation and improve the precision of predicting LAI. Itfound that reflectance in894.9nm waveband have a direct relation to the LAI, thecubic polynomial regression model is the best model and the correlation coefficient is0.66, its relative error is7.49percent. However it found that the wavelength of bestcorrelation using the reflectance spectrums that have been derivatived is732.1nm, and the model is quadratic polynomial regression, the correlation coefficient is0.68andrelative error is4.169percent.There are five indictors are used to estimate vegetation index. The bestwavebands are selected using MATLAB, the wavelength of the NDVI、RVI、MSR、SAVI、MSAVI are752.4nm、701.5nm、823.7nm and701.5nm. It found by all linearregression models that the exponential regression model with SAVI as theindependent variable (Y=1.171e1.064x) has highest estimation accuracy, and thecorrelation coefficient is0.674and relative error is2.901percent. Found by themultiple linear regression model which using multivariate stepwise regression: Thecumulative contribution rate of the first two principal components is99.064percentafter principal component analysis, achieve dimensionality reduction, at the same time,only loss little amount of information of the original data, the multivariate stepwiseregression model by them as the alternative variable is Y=2.796+0.052*F1+0.04*F2((F1) is the fist principal component,(F2) is the second principal component), itsrelative error is3.123percent, less than the linear regression models.Found by different types of statistical regression models, these models havehighest estimation accuracy is90percent in all models, the exponential regressionmodel with SAVI is the best model, using the model to invert maize leaf area index ofthe area and analysis its growth and space distribution.It is feasible to quantitative estimate the LAI values in a wide range fast andaccurately using Hyperspectral remote sensing image. The study about monitoringfood crops using remote sensing in theories and applications is promoted, and thebasic theories of monitoring and assessing maize are supplied by it.