Integrated hyperspectral monitoring of amylose content across multi-processing morphotypes in broomcorn millet (Panicum miliaceum L.)

Amylose content (AC) is a key determinant of the processing quality and genetic attributes of broomcorn millet ( Panicum miliaceum L. ). In this study, 134 broomcorn millet accessions from both domestic and international sources were systematically evaluated over two consecutive years (2022–2023) to characterize spectral differences among various processing forms (grain, millet, and flour) and grain colors, and to explore their relationships with AC. High-precision hyperspectral prediction models were developed and validated. The results showed that white-grained accessions exhibited the highest reflectance in the visible (400–780 nm) and near-infrared (1100–2450 nm) regions, whereas black-grained accessions showed the strongest absorption in the near-infrared region. Correlation analysis revealed a significant positive association between thousand-grain weight and AC (r = 0.211), while the length-to-width ratio was strongly and negatively correlated with yield (P < 0.01), suggesting that rounder grains are more conducive to high-yield breeding. Among the modeling approaches, the modified chlorophyll absorption ratio index (MCARI) combined with Savitzky–Golay (SG) preprocessing achieved the highest prediction accuracy for millet (validation set R 2 = 0.858, RPD = 2.0). The full-spectrum partial least squares (Full-PLS) model demonstrated the best overall predictive performance, with validation R 2 values of 0.615, 0.897, and 0.923 for grain, millet, and flour, respectively. This study provides an efficient and non-destructive approach for quality assessment and precision breeding in broomcorn millet, with potential applications in crop phenomics and digital agriculture. • The first three-element morphological spectral library of broomcorn millet was created. • The synergistic enhancement mechanism of round grains was discovered. • A Full-R-PLS universal model was constructed, with an R 2 of 0.923 in the flour validation set. • The MCARI-SG optimization algorithm was developed, achieving precise field prediction.