Abstract:In order to detect leaf chlorophyll level nondestructively and instantly. VIS/NIR reflection spectroscopy technique was examined. In the test. 70 leaf samples were collected for model calibration and another 50 for model verification. Each leaf sample was optically measured by USB4000, a modular spectrometer. By the observation of spectral curves, the spectral range between 650nm and 750nm was found significant For mathematic modeling, of leaf chlorophyll level. SPAD-502 meter Was used for chemometrical measurement of leaf chlorophyll value. In the test, it was found necessary to Put leaf thickness into consideration. The procedure of shaping the prediction model is as follows: First, leaf chlorophyll level prediction equation was created with uncertain parameters. Second, a genetic algorithm was programmed by Visual Basic 6.0 for parameter optimization. As the result of the calculation, the optimal spectral range was narrowed within 683.24nm and 733.91nm. Compared with the R2=0.2309 for calibration set and R2=0.5675 for verification set without concerns of leaf thickness, the effect of leaf thickness on the spectral modeling is significant: the R2 of calibration set and verification set has been improved as high as 0.8658 and 0.9161 respectively. The test showed that it is practical to use VIS/NIR reflection spectrometer for the quantitative determination of leaf chlorophyll level.

Application Of Genetic Algorithm In The Modeling Of Leaf Chlorophyll Level Based On Vis/Nir Reflection Spectroscopy