Identification of Sicilian landraces and Canadian cultivars of lentil using an image analysis system

Abstract

Seed colour and size determine the appearance of grain legumes such as lentils (Lens culinaris Medik.). Currently, these characteristics are assessed by visual inspection of grain; it is slow and subjective, for this reason an objective imaging system was developed in Canada to measure bulk seeds lentil colour and size in a consistent manner for lentil quality colour grading. The Canadian imaging method, using image captured by a flatbed scanner, was modified in collaboration with the Stazione Sperimentale di Granicoltura (Research Institute in Sicily) to measure seed size, shape and mean colour on individual seeds; each seed imaging data were computed with a Linear Discriminant Analysis (Classifier) to identify five Sicilian landraces of Lens culinaris Medik., precisely Aragona, Bronte, Leonforte, Ustica and Villalba and three common Canadian accessions, cv Laird, cv Crimson and cv Eston. The performance of the classifier was 99.8% for the training sets and 97.1% for the independent test set. In addition to commercial international trade, lentil seed characterisation is very important to identify and catalogue in a biodiversity conservation program.