Detection of impurities on faba bean (Vicia faba L.) by NIR spectroscopy

Faba bean (Vicia faba L.) holds a significant position as one of the primary agricultural commodities globally. The presence of impurities on faba beans can lead to significant economic losses and quality deterioration that influences seed vigor and growth. Therefore, it is crucial to detect impurities seeds rapidly and non-destructively. In this study, a near-infrared (NIR) spectra acquisition device (400–1000 nm) was employed for seed quality detection. Spectral fingerprints extracted from pure faba bean seed and impurities were modeled using principal component analysis (PCA), partial least square (PLS) regression and linear discriminant analysis (LDA) to demonstrate the general overview of the spectral characteristics, predict the seed and impurities features and classify the seeds and impurities to the right categories. The results showed that impurities can be detected and classified precisely with total explained variance of 100%, with better separation of the classes. It also indicates that good statistics were obtained for prediction, cross-validation, and calibration, the PLS model achieved correlation coefficients (r) of 0.97, with minimal values of RMSE of about 2.98. LDA was utilized to classify the seeds based on their spectral fingerprints, achieving an overall classification accuracy of 84%. The model effectively distinguished between pure seeds and impurities, demonstrating its potential for rapid, non-destructive impurity detection in faba bean seeds. This study illustrates the applicability of NIR spectroscopy combined with PCA, PLS, and LDA models for accurate seed impurity detection and classification.