Selecting optimal hyperspectral method for estimation of biochemical concentration of FCV tobacco leaf at the maturity stage

The goal of this work is to develop non-destructive techniques that can conveniently, rapidly and accurately assess FCV tobacco leaf biochemistry status at the maturity stage at leaf levels by selecting optimal hyperspectral method for estimation of biochemistry concentration and biophysical parameter. The data of reflectance and absorption spectra of fresh tobacco leaf were measured with ASD Fieldspec FR2500 at the maturity stage, and the parameters of biophysical and biochemistry of the leaf were collected by normal method. The correlation between spectral parameters consisting of single band reflectance, logarithm of reciprocal of reflectance, transmittance and absorption and 100 existing spectral indices selected in the visible and near-infrared spectrum(350～1650 nm), with biochemistry concentration was analyzed by using linear, power and exponential fitting. Optimal method was selected based on the strongest strength of determination coefficient. The result indicated that optimal optical methods for chlorophyll a, chlorophyll b, carotenoids and potassium concentration of leaf were respectively Gitelson and Merzlyak２(GM₂)(R²=0.81), Photochemical Reflectance Index 2(PRI₂)(R²=0.80), Gitelson and Merzlyak２(GM₂)(R²=0.83), the starting position of reflectance gallery at 1420 nm(λs₁₄₂₀)(R²=0.67) using linear fitting. Water-solubility total sugar, specific leaf weight(SLW), total nitrogen, nicotine were respectively first derivative of reflectance at 532 nm using linear fitting(R²=0.54), transmittance at 1423 nm using linear fitting(R²=0.45), first derivative of logarithm of reciprocal of reflectance (δlog(1/R)) at 666 nm using power fitting(R²=0.44), second derivative of logarithm of reciprocal of reflectance (δδlog(1/R)) at 1135 nm using power fitting(R²=0.20). It is feasible to predict matter concentration of tobacco leaf by optimal optical method for spectral parameters.