Computer vision as a method complementary to molecular analysis: grapevine cultivar seeds case study.

Despite different breeding events, as well as the domestication phenomena which contributed to enrich the grape varietal heritage in Sardinia, many local varieties simply are the product of linguistic distorting due to the wide heterogeneity historic-cultural of the island. This phenomenon generated a great assortment of grape names, that, together with the huge real number of cultivars, is the cause of the incredible current grapevine Sardinian panorama. The goal of this article is to compare the published molecular data of 40 Sardinian autochthonous cultivars with the results achieved by the germplasm phenotypical characterization, on the basis of morpho-colorimetric features and Elliptic Fourier Descriptors (EFDs), measured by image analysis. Statistical classifiers were implemented to discriminate dissimilar seeds and carry out hypothetical synonymy groups to compare with those proposed on the basis of Simple Sequence Repeat (SSR) markers. This work represents the first trial to validate a morpho-colorimetric characterization method by direct comparison with molecular data, proving that the 113 measured features of the germplasm resulted adequate to achieve a clear discrimination among the synonymy groups.