Study On Modeling Total Nitrogen And Posphorus Content Of Korla Fragrant Pear Leaves Based On Hyperspectrum By Analysis

This paper based on field experiments and regional samling of Korla Fragrant Pear tree.Hyperspectral reflectance were measured by SVC HR-768in leaves of20-years Korla FragrantPear tree, analyzed nitrogen concentration and phosphorus concentration which in differentnitrogen rate and phosphorus rate. Correlation between characteristic spectral parameters andnitrogen concentration phosphorus concentration of leaves was analyzed to determine sensitivebands which prediction models were built using stepwise regression method. Finally, thepredictive precisions are analyzed for some chosen models in order to determine the bestestimation models. The results are following:(1) The hyperspectral reflectance of pear leaves was markedly different under variednitrogen and phosphorus concents, the response of leaves reflectance spectra to the nitrogen inthe visible wavebands and NIR wavebands were obvioused. In the NIR wavebands the reflectancetended to indecreasing with increasing nitrogen nitrogen; The response of leaves reflectancespectra to the phosphorus in the NIR wavebands (760～1000nm) were obvioused, whichreflectance tended to indecreasing with increasing phosphorus concents.(2) The Pear leaves’ nitrogen contents with the raw hyperspectral reflectance of sensitivebands at720nm, with the first derivative hyperspectral reflectance of sensitive bands at720nm,with Dr and SDr and Rg/Ro and (Rg-Ro)/(Rg+Ro) reached the significant correlation,correlation coefficient being more than0.809;Leaves’phosphorus contents with the rawhyperspectral reflectance of sensitive bands at757nm, with the first derivative hyperspectralreflectance of sensitive bands at425nm and715nm, with Dr and SDr reached the significantcorrelation, correlation coefficient be-0.972and-0.984.(3) Linear model using stepwise regression method, leaves’ nitrogen and phosphorus contentswith raw hyperspectral reflectance and the first derivative hyperspectral reflectance of sensitivebands, and hyperspectral characteristic which reached the significant correlation. Uing the relativeerror range, the average relative error and the decision coefficient of three indicators to verify theaccuracy and stability of the model. The last,The first derivative reflectance at703nm wasdetermined as the best model for estimating nitrogen concentration of Korla Fragrant Pear leaves,Y=50.535X703-40.58the SDr was determined as the best model for estimating phosphorusconcentration of Korla Fragrant Pear leaves Y=-0.229X+2.411...