Summary
This controlled laboratory study demonstrates the application of visible–near-infrared–shortwave infrared hyperspectral imaging (449.54–2399.17 nm) to non-destructively classify bean seed accessions, addressing limitations of conventional destructive assays for germplasm conservation and breeding. Linear discriminant analysis achieved 96.35% balanced accuracy on full-spectrum data, substantially outperforming reduced-band configurations and deep learning approaches. The work establishes a methodological benchmark for scalable, non-destructive seed identification in legume breeding programmes.
UK applicability
The methodology could support UK germplasm conservation initiatives and breeding programmes for pulses and legumes, enabling non-destructive, high-throughput seed accession identification. However, applicability depends on access to hyperspectral imaging equipment and trained personnel, and performance under field or variable conditions remains untested.
Key measures
Balanced accuracy, macro-F1 score, precision, recall, R² values at specific wavelengths (461.37 nm), principal component analysis variance explained, test accuracy
Outcomes reported
The study evaluated visible–near-infrared–shortwave infrared hyperspectral imaging combined with machine learning algorithms to classify 32 grain-legume accessions (30 common bean landraces and 2 outgroup legumes) based on spectral signatures. Performance metrics included balanced accuracy, macro-F1 scores, precision and recall across multiple classification approaches.
Topic tags
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.