Study On Chlorophyll And LAI Of Corn Based On Hyperspectral Remote Sensing Model

In this paper, we analysis the corn physiological parameters, such as LAI andchlorophyll content with high spectral inversion model based on the measured cornhyperspectral data. The measured data and the environmental satellite hyperspectralimage data are used to monitor the corn growing of central and western regions ofJilin Province. The reflectance of corn, first derivative of spectra, spectral vegetationindex and partial least squares method are used to establish the statistical regressionmodels of estimating corn physiological parameters, LAI and chlorophyll content.Those methods were used and compared for extracting hyperspectral information andalso improving the accuracy of estimating vegetation biophysical parameters. At thesame time, we use three band model, which is a semi-analytical model, to estimate theleaf chlorophyll content of corn. Using the methods mentioned above to analysis leafarea index and chlorophyll content of corn, we can get the following conclusions:1Throughout the growing seasons, the correlation coefficient of the spectralreflectance and leaf area index in the visible band shows a strong negativecorrelation, and there is a sharp rise in the red. But the correlation coefficientwhich is high negative correlation first and then converts into a strong positivecorrelation in the near infrared and the correlation is relatively stable, most ofwhich is the positive correlation. The overall correlation coefficient of thederivative spectra and leaf area index is higher than that of the reflectance and leafarea index. As far as chlorophyll content is concerned, throughout the growingseason, correlation coefficient of the spectral reflectance and chlorophyll isrelatively stable in the entire wavelength range, which is showing a negative correlation. The volatility of correlation coefficient, between first derivative ofspectrum and chlorophyll content, is relatively large in the entire wavelengthrange and bigger than that of reflectance and chlorophyll content. Similar to thevolatility of correlation between first derivative of spectra and LAI, volatility ofcorrelation with chlorophyll content is relatively large in the entire wavelengthrange, and it also has a number of maxima and minima.2This study aims to estimate the effect of a variety of traditional vegetation indexestimating LAI and chlorophyll content of corn. Six kinds of vegetation index,NDVI, DVI, RVI, RDVI, PVI and SAVI, are used for the revision of LAI, inwhich the result based on NDVI is the best and the most stable while the accuracybased on RVI is the worst. And estimating chlorophyll content are used our kindsof vegetation index, such as NDVI, MSR, MCAVI/OSAVI and TCAVI/OSAVI.In addition, the soil line due to the Perpendicular Vegetation Index (PVI) can bebetter to filter out the influence of soil background and with PVI high retrievalaccuracy can be achieved. The remaining three models in the inversion of LAI didnot show the great advantage. Among vegetation index in estimating chlorophyll,the result which dues to NDVI is most desirable. And accuracies based on MSRand TCAVI/OSAVI is followed. The model established by MCAVI/OSAVI getsthe worst result.3The high-spectral reflectance spectral data which exist with more independentvariables and potential influent factors can be solved with the method of Partialleast squares. It can remove samples not suitable easily from the residuals of thecalibration model and build the best model quickly. The inversion of LAI andchlorophyll content based on the PLS method can get ideal results. LAI inversionaccuracy based on PLS is more or less with the result based on vegetation index(NDVI), while the precision of chlorophyll content inversion is higher comparedwith the accuracy based on vegetation index.4Three-band model in the inversion of chlorophyll content can remove theinterference of other colors, background noise and other factors, so in this studywe obtain the desired results (R~2=0.703). And this method is a feasible and desirable way in estimating chlorophyll content. The result based on Three-bandmodel is higher than that based on VIs and single band but a little lower than theaccuracy of PLS.5The results of using vegetation index methods in the inversion of LAI are good,based on HSI image. The coefficients of determination are between0.5and0.8.NDVI is the highest accuracy (R~2=0.773) and the DVI is the lowest (R~2=0.506).While estimating the chlorophyll content, the coefficient of determination of themodel in the modeling is not high. But the verification accuracy is rather high(R~2>0.9). And the slope of validated models is less than0.4, which indicates thatestimates are lower than the measured results.