Discrimination of 14 olive cultivars using morphological analysis and machine learning algorithms.

Area of Science:

• Agricultural Science
• Genetics
• Plant Biology

Background:

• Traditional olive germplasm identification relies on subjective, labor-intensive morphological measurements.
• Automated quantification of geometrical features offers higher accuracy and efficiency.

Purpose of the Study:

• To develop and validate an automated phenomics methodology for olive cultivar identification and discrimination.
• To assess the efficiency of machine learning algorithms in classifying olive cultivars based on morphological data.

Main Methods:

• Quantification of 24 fruit, 16 leaf, and 25 endocarp characteristics using automated methods.
• Application of basic classifiers combined with a meta-classifier approach based on machine learning.
• Analysis of data from two consecutive olive growing seasons.

Main Results:

• Discrimination of 14 olive cultivars with 95% accuracy using the meta-classifier approach.
• Quantitative assessment of the contribution and significance of each morphological feature for cultivar discrimination.
• Identification of key features for discriminating most cultivars (endocarps and fruits), with leaf features specific to the Kalamon cultivar.

Conclusions:

• The proposed automated phenomics methodology is an efficient tool for olive cultivar identification and discrimination.
• Combining morphological features from multiple olive organs enhances discrimination capacity.
• This approach has broad applications in olive breeding and germplasm management.