Hot air drying is widely used for wheat due to simple operation, but research on the intelligent regulation of the drying system based on product quality is limited. This study explored the application of hyperspectral imaging (HSI) for non-contact detection of wheat quality after hot air drying. Wheat characteristics after hot air drying were first analyzed at different drying treatments. Chemometric methods were combined with HSI for regression modeling of the wheat drying quality. At air speeds of 2.0, 1.5, and 1.0 m/s, the drying times at 40 °C, 60 °C, and 80 °C were 110–130 min, 40–50 min, and 20–30 min, respectively. After drying, wheat exhibited higher L⁎ (58.18–60.77) and a⁎ values (6.74–7.69), and ΔH values (3.91–6.05 J/g), with changes in microstructure. Within the spectral range of 400–1000 nm, the regression capabilities of three models were compared based on raw spectral data for various physicochemical indicators. The PLSR, with test R2 ranging 0.93–0.99, was selected to perform regression of quality indicators based on preprocessed spectral data. The results showed that the first and second derivative-treated spectral data exhibited good performance when regressed by the PLSR (test R2 > 0.98 with RPD > 2.5). This study demonstrates that combining HSI with machine learning exhibits favorable regression performance for wheat quality, facilitating its widespread application in quality detection during hot air drying.

Paper Linkage:https://doi.org/10.1016/j.foodres.2026.119175