Hyperspectral Inversion of Chlorophyll Content in Maize Leaves Based on Image and Spectrum Fusion

Abstract

Chlorophyll content is an important indicator for evaluating the growth status of maize crops. To make full use of hyperspectral information for the prediction of chlorophyll content in maize leaves, this study analyzed the correlation between Gray-level Co-occurrence Matrix (GLCM) features and chlorophyll content, and used Competitive Adaptive Reweighted Sampling (CARS) and Random Frog (RF) to select the sensitive wavelengths of the spectra, and establish a partial least squares regression (PLSR) model for predicting chlorophyll content in maize leaves after fusion of spectral features and image texture features. The results show that the PLSR model constructed by combining the GLCM features with the spectral features selected by CARS has the best effect, and the prediction accuracy is effectively improved compared with that before the fusion of GLCM features, the coefficient of determination R2 is improved from 0.8802 to 0.9104, and the relative analysis error RPD is improved from 3.1938 to 3.6166. Therefore, the PLSR regression model based on GLCM and spectral features can effectively predict the chlorophyll content of maize leaves.