Leveraging deep learning and spectral information for enhanced variety identification of safflower seeds.

To achieve rapid and non-destructive identification of safflower seeds, a qualitative identification model was constructed based on near-infrared spectroscopy (NIRS) combined with chemometric methods including principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), backpropagation neural networks (BPNN), and convolutional neural network (CNN). Results: The model utilizing raw safflower seed spectra combined with BPNN demonstrated the highest performance in variety differentiation, achieving 100% prediction accuracy on both the training and test sets. The model combining raw spectra with CNN ranked second, with prediction accuracies of 100% and 97.235% respectively. The synergistic integration of neural networks' deep learning prowess with the inherent benefits of near-infrared spectroscopy establishes a robust analytical framework for safflower seed germplasm authentication. This methodology presents substantial potential as an advanced approach for provenance verification and quality surveillance in both food and pharmaceutical sectors.