Multispectral Imaging for Discriminating Common Bean Seeds under Contrasting Vigor Conditions Induced by Accelerated Aging

Background: Accurate assessment of seed vigour is essential for ensuring the physiological quality of common bean (Phaseolus vulgaris L.) seeds. Conventional methods may be time-consuming and involve destructive procedures. Multispectral imaging offers a rapid, objective and non-destructive alternative. However, the relative effectiveness of morphological, shape, colour and spectral features in discriminating common bean seeds subjected to accelerated ageing remains insufficiently investigated. Aims: To evaluate whether multispectral imaging can discriminate common bean (Phaseolus vulgaris L.) seeds under contrasting vigour conditions, and to compare the discriminant capacity of morphological, shape, colour and spectral features. Study Design: Completely randomised design with two experimental conditions (high vigour and low vigour induced by accelerated ageing) and four replicates of 50 seeds. Place and Duration of Study: Seed Analysis Laboratory, Department of Agricultural Sciences, State University of Montes Claros (UNIMONTES), Janaúba, Minas Gerais, Brazil, in 2026. Methodology: Seeds of the cultivar Pérola from a single lot were either kept unaged (high vigour) or aged at 41 °C for 72 h (low vigour). Images were acquired with a VideometerLab4 at 19 wavelengths (365–970 nm), and 35 morphological, shape, colour and spectral reflectance variables were extracted per seed. Data were analysed by Welch's t-test on replicate means, principal component analysis and linear discriminant analysis (LDA) with leave-one-replicate-out cross-validation. Results: Five of the six morphological variables, the three CIE L*a*b* colour components and 17 of the 19 spectral bands differed between conditions. The first principal component, dominated by spectral variables, explained 47.5% of the total variance. LDA accuracies were 26.4%, 63.2% and 68.5% for shape, morphological and colour variables, respectively. The 19 spectral bands achieved the highest accuracy (98.2%), correctly classifying 197 of 200 high-vigour seeds and 190 of 194 low-vigour seeds; combining all 35 variables yielded 98.0%, with no gain over spectral information alone. Conclusion: The spectral response concentrates a high capacity for discrimination between the experimental conditions evaluated, showing the potential of multispectral imaging as a rapid, objective and non-destructive tool for discriminating common bean seeds subjected to accelerated ageing. Validation with commercial lots that differ naturally in vigour is still required.