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

Area of Science:

• Viticulture and Oenology
• Plant Genetics
• Agricultural Science

Background:

• Sardinia possesses a rich grapevine heritage, complicated by linguistic variations and numerous local cultivars.
• Distinguishing between similar grape varieties is challenging due to historical naming conventions and genetic diversity.
• Accurate characterization is crucial for preserving and managing this unique viticultural landscape.

Purpose of the Study:

• To compare molecular data with phenotypical characterization of 40 Sardinian autochthonous grapevines.
• To validate a morpho-colorimetric characterization method using image analysis and statistical classifiers.
• To assess the effectiveness of 113 measured germplasm features in discriminating grapevine synonymy groups.

Main Methods:

• Phenotypical characterization of grapevine germplasm using morpho-colorimetric features and Elliptic Fourier Descriptors (EFDs) via image analysis.
• Application of statistical classifiers to discriminate between dissimilar seeds and identify potential synonymy groups.
• Comparison of phenotypical results with existing Simple Sequence Repeat (SSR) marker data.

Main Results:

• The study successfully discriminated between dissimilar seeds using 113 measured germplasm features.
• Morpho-colorimetric characterization, analyzed with statistical classifiers, proved adequate for distinguishing grapevine synonymy groups.
• The phenotypical characterization results showed good agreement with molecular data (SSR markers).

Conclusions:

• This work validates morpho-colorimetric characterization as a reliable method for grapevine discrimination in Sardinia.
• The 113 measured features are sufficient to clearly differentiate between grapevine cultivars and synonymy groups.
• This approach aids in understanding and preserving Sardinia's complex and diverse grapevine genetic resources.