Olive Variety Classification and Prediction From 3D Morphology of Fruit and Stone: A Study Case on Five South Italy Autochthone Cultivars.

Abstract

Accurate olive cultivar identification is critical for ensuring quality control and traceability in the olive oil industry. The International Olive Council (IOC) and the International Union for the Protection of New Varieties of Plants (UPOV) have established standardized protocols for varietal characterization. Over the past two decades, two-dimensional image analysis techniques have been increasingly employed for olive variety identification, utilizing various morphological parameters and machine learning approaches. This study investigates olive varietal classification through three-dimensional morphological analysis of fruits and stones using X-ray microtomography. The research evaluates the discriminative power of different trait combinations using both Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) algorithms to contribute to an optimized protocol for cultivar identification. Five autochthonous olive cultivars from the Campania region (Southern Italy) were analyzed. A preliminary comparison of classification performance between continuous and discrete morphological olive data revealed superior effectiveness of the continuous ones. Integrating quantitative morphometric traits with selected visual discrete UPOV characteristics yielded optimal overall classification accuracy of 88.41% using LDA with 84.4% for Ravece, 81.5% for Ortice, 100% for Frantoio, 81.3% for Rotondella, and 90.9% for Minucciola olive varieties. The best variety prediction rates, based on an olive sample not used for training, were provided by SVM, obtaining 70.0% for Ravece, 87.5% for Ortice, 54.5% for Frantoio, 60.0% for Rotondella, and 66.7% for Minucciola. Quantification of varietal overlap through Bhattacharyya coefficients identified Ortice and Ravece as the most phenotypically similar varieties, while Rotondella and Minucciola exhibited the most distinctive fruit morphology. Notably, all varieties showed at least one misclassification with the Frantoio variety. Morphological analysis demonstrated that endocarp surface traits provided the most discriminative power, and internal cavity characteristics also contributed significantly to varietal differentiation. These findings suggest two key implications: potential updates of UPOV guidelines for distinctness evaluation protocols and promising applications in authenticity verification for high-quality olive products.

Plant phenotyping relevance

X線マイクロトモグラフィーによる果実・核の3次元形態計測と、形態形質を用いた分類性能評価が研究の中心であり、品種識別用の植物表現型取得・解析手法に該当する。

Code and data availability

The paper describes olive micro-CT phenotyping with LDA/SVM analysis, but no public deposit of the phenotype datasets, micro-CT images, or analysis code is provided. Supplementary Files S1/S2 are referenced without public URLs. The scikit-learn link is a generic library, the Bruker link is equipment documentation, and

No evidence-backed public reproduction asset is currently recorded.