Laser-induced fluorescence spectroscopy combined with multivariate analysis for rice seeds and grains discrimination

Rice is a staple food in sub-Saharan Africa, including Ghana. Local production is hindered by the use of rice grains for cultivation, which directly affects both yield and grain quality. This study employed laser-induced fluorescence spectroscopy (LIFS) combined with multivariate analysis to rapidly and non-destructively discriminate between rice seeds and grains, offering an alternative to conventional methods. Fluorescence spectra from rice seeds and grains from six locally cultivated rice varieties were analyzed using three pre-processing techniques (Z-score, first derivative, and second derivative) across four machine learning models: linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), K-nearest neighbor (KNN), and support vector machine (SVM). Principal component analysis score plots and the Davies-Bouldin index were used to assess the separation between seeds and grains. The first and second derivatives outperformed the Z-score, with SVM performing well under the second derivative and KNN excelling under the first derivative. LDA and QDA varied depending on the rice variety and the pre-processing method. The best accuracy was achieved using the first derivative with KNN, achieving test accuracy and F1 scores ranging from 0.74 to 0.97 and 0.76 to 0.97, respectively, confirming that LIFS combined with multivariate techniques is an effective method for discriminating rice seeds and grains.