A segmentation approach for the identification of hazelnut cultivars

Abstract

The development of a methodology for cultivar discrimination and selection during on-line quality selection could be a relevant tool for certain industries that use hazelnuts as ingredients in processed food products. The aim of this work was to obtain a computer vision system that enables an automatic recognition of six hazelnuts cultivars from 3 different origins (US, Italy and Spain). To achieve this objective, an algorithm for tuning the segmentation between the hazelnut shell and crown was developed. Important morphological features that allow the correct identification of the cultivar present in an image were obtained from the segmentation process. Finally, these features were used as input data to a neural network classifier. The obtained results showed a high percentage of correct cultivar classification.