Adaptive spectral band optimization for low-cost multispectral estimation of maize leaf chlorophyll using a LASSO–IARO strategy

Abstract

Chlorophyll content in maize leaves is an important indicator of the physiological status and nutritional conditions of the crop. Rapid and cost-effective monitoring of chlorophyll is therefore essential for precision agriculture. However, spectral measurement instruments are expensive and difficult to deploy widely in field environments. In practical applications, low-cost multispectral sensors require the selection of a small number of informative spectral bands while maintaining prediction accuracy. In this study, an adaptive spectral band optimization strategy integrating the least absolute shrinkage and selection operator (LASSO) and an improved artificial rabbits optimization (IARO) was proposed to identify compact band subsets for chlorophyll estimation. The spectral data of maize leaves were preprocessed using Savitzky–Golay smoothing (SG) and Standard Normal Variate Transformation (SNV). Feature bands were selected using SPA, Pearson correlation, LASSO, and the proposed LASSO–IARO method, and predictive models were developed using partial least squares regression (PLSR) and support vector regression (SVR). Results showed that LASSO–IARO reduced the number of selected bands by 42.86–73.33% compared with conventional methods while maintaining comparable prediction accuracy. The LASSO–IARO–PLSR model achieved a coefficient of determination (R²) of 0.81 with a root mean square error (RMSE) value of 2.01 on the testing set. The optimized band subset (517 nm, 520 nm, 696 nm, and 730 nm) provides candidate wavelengths for designing low-cost multispectral sensors for in-field chlorophyll monitoring.