Automatic plant phenotyping analysis of Melon (Cucumis melo L.) germplasm resources using deep learning methods and...

Area of Science:

• Horticultural science
• Plant breeding
• Computer vision

Background:

• Melon (Cucumis melo L.) breeding requires efficient phenotypic analysis.
• Current manual methods are time-consuming, costly, and subjective.
• Automated analysis is needed to accelerate melon germplasm selection.

Purpose of the Study:

• To develop an AI-driven framework for automated melon phenotypic trait extraction.
• To validate the accuracy and efficiency of the AI system compared to manual measurements.
• To create a user-friendly tool for rapid melon phenotype acquisition.

Main Methods:

• Utilized artificial intelligence (AI) with deep learning models: Dual Attention Network (DANet), RTMDet, RTMPose, and MobileSAM.
• Processed images of 117 melon varieties over two years.
• Developed feature extraction algorithms to obtain 11 phenotypic traits.

Main Results:

• The AI framework accurately segmented melon fruit and pedicel.
• High correlation observed between algorithm-predicted and manually measured phenotypic traits.
• Cluster analysis showed strong consistency between classification results and melon genotypes.

Conclusions:

• The developed AI system provides an efficient and accurate tool for melon phenotypic analysis.
• This technology accelerates melon breeding programs and genotype-phenotype research.
• Automated phenotyping enhances the selection of superior melon germplasm resources.