Research on corn quality detection based on near-infrared spectroscopy

To improve the efficiency and accuracy of corn quality detection, this study aims to utilize near-infrared (NIR) spectroscopy for the nondestructive analysis of key quality components in corn, including moisture, oil, protein, and starch. By applying first derivative (FD) preprocessing to the spectral data, background noise was removed, and signal features were enhanced. Multiple dimensionality reduction algorithms, such as synergy interval partial least squares (SiPLS), competitive adaptive reweighted sampling (CARS), successive projections algorithm (SPA), and correlation coefficient (CC), were employed to select the most representative characteristic wavelengths. Predictive models were then established using support vector machine (SVM) and backpropagation (BP) neural network. The results demonstrated that the SiPLS-CARS-SVM model exhibited superior performance in predicting moisture, oil, protein, and starch content, with higher coefficients of determination (R2P), lower root mean square error of prediction (RMSEP), and higher residual predictive deviation (RPD) compared to the BP neural network model. This study shows that NIR spectroscopy combined with the SiPLS-CARS-SVM model can significantly enhance the efficiency and accuracy of corn quality detection, providing a valuable reference for corn quality assessment and production.