Identification of Italian landraces of bean (Phaseolus vulgaris L.) using an image analysis system

Colour, shape and size of whole seeds and their spots of some Italian landraces of bean (Phaseolus vulgaris L.) were measured using a specifically developed macro, based on image analysis library KS-400 V 3.0 (Carl Zeiss, Germany). Statistical classifiers to identify 15 landraces, one landrace for two cultivation regions and two landraces for different harvest years, were obtained using the image analysis data elaborated with the Linear Discriminant Analysis algorithm. The macro, called Bean.mcr, works quickly and accurately. The performance of the four classifiers was as follows: the first, for the identification of 15 landraces, 99% training set and 98.20% test set, respectively; the second, to identify two harvest years, 99% and 100%; the third, to identify three harvest years, 100% and 98.82%: the fourth, for the identification of the two different cultivation regions, 94.5% and 82.40%. The macro gives, to a certain extent, the possibility of tracing beans yield, in order to get a.. market card" for landrace beans. (c) 2009 Elsevier B.V. All rights reserved.