From Laboratory Accuracy to Industrial Deployment: A Review of Hyperspectral Imaging for Cereal Grain Quality and Safety Assessment

Hyperspectral imaging (HSI) is a powerful non-destructive tool for cereal grain quality and safety assessment, enabling simultaneous prediction of multiple quality parameters through integrated spectroscopy and spatial imaging. While conventional chemometric methods such as partial least squares regression remain widely used, deep learning, particularly convolutional neural networks and ensemble methods, has demonstrated superior performance and is increasingly adopted. This review synthesizes HSI applications across six major cereal crops (maize, rice, wheat, barley, sorghum, and millet), covering nutritional composition, moisture, mycotoxins, physical traits, and variety classification. Spectral principles, modeling approaches, and cross-crop patterns are analyzed. Critical gaps are identified, including the near-absence of lipid and dietary fiber studies and scarce industrial deployment validation. Publication bias and methodological heterogeneity across primary studies, which likely inflate reported prediction accuracies, are discussed. This review provides a comprehensive reference and offers perspectives on advancing HSI toward standardized and industrially deployable solutions.