Non destructive parameters by CVS for predictive modelling and machine learning in Salicornia europaea

Purpose

The study aims to develop and validate a low-cost, non-destructive method using a computer vision system (CVS) combined with multivariate analysis to classify Salicornia europaea plants based on morphometric and colour traits. The ultimate goal is to support selective breeding by predicting plant biomass and substrate salinity, enabling effective differentiation of salt tolerance levels.

Nature

This is an experimental and analytical study involving digital image capture, quantitative trait analysis, and predictive modelling. The method integrates morphometric measurements and CIELab-based colour metrics extracted via CVS, followed by statistical techniques such as Pearson correlation, PCA, MDA, and multiple linear regression to classify and predict phenotypic and environmental parameters.

Scope

The study evaluated 120 plants from two distinct S. europaea populations under varying salinity levels. It demonstrated high prediction accuracy for biomass and substrate salinity across different tolerance levels, with up to 100% validation accuracy in classification. The approach proves scalable and applicable to ecological monitoring, bio-agriculture, and industrial halophyte cultivation, with potential for integration into AI and mobile platforms.