ESTIMATION OF CHLOROPHYLL CONTENT IN APPLE LEAVES BASED ON IMAGING SPECTROSCOPY

Abstract

To promote the use of imaging spectroscopy to assess the nutritional status of apple trees, the models to estimate the chlorophyll content of apple leaves were explored. Spectral data for apple leaves were collected with an imaging spectrometer and then preprocessed with the nine-point moving weighted average method. Correlation analyses were conducted between chlorophyll content and mathematically transformed spectral data. Wavelengths sensitive to chlorophyll content were selected on the basis of the highest correlation coefficients, and partial least squares (PLS), support vector machine (SVM), and random forest (RF) models to estimate chlorophyll content were established and tested. The wavelengths sensitive to chlorophyll content were 414, 424, 429, 439, and 577 nm. The best model was the SVM model with wavelength data subjected to a second order differential of the logarithm transformation (lgR₄₁₄)², (lgR₄₂₄)², (lgR₄₂₉)², (lgR₄₃₉)², (lgR₅₇₇)² as the independent variables. For this model, the coefficient of determination V-R² was 0.7372, the root mean square error V-RMSE was 0.4477, and the residual predictive deviation V-RPD was 1.8810. Among all the models, this SVM model had the highest V-R² and V-RPD values and the lowest V-RMSE value.