Abstract
The scope of this work is to propose a method based on near infrared hyperspectral imaging to automatically sort different parts of onions (dry peel, outer skin and onion bulbs) produced during specific industrial processes. For this purpose, two procedures have been developed; one is based on computation of the spectral area related to the organic sulphur band and the second one based on the discriminant chemometric tool PLS-DA. Both procedures have been compared by repeated measurements on artificial mixtures of dry peel and onions. This comparison indicated that PLS-DA produced the most accurate values while the study of spectral area seems to have the best repeatability. The PLS-DA procedure was then applied to several samples of mixed dried onions to compare the results with reference values obtained by manual counting. Based on the measurement repeatability, the stability of the results and the small differences between the reference and predicted values, this study has demonstrated the ability of near infrared hyperspectral imaging to identify and quantify the different parts of dry onions in mixtures.
Get full access to this article
View all access options for this article.