Organ-level 3D phenotyping of saffron using a low-cost dual-camera workflow

Area of Science:

• Plant Biology
• Agricultural Science
• Computational Biology

Background:

• Precise, non-destructive 3D phenotyping of saffron is crucial for yield optimization and breeding.
• Extreme leaf morphology (narrow, overlapping, prostrate) challenges traditional imaging, limiting analysis to whole-canopy descriptors.
• A novel workflow using a low-cost dual-camera system and Structure-from-Motion Multi-View Stereo (SfM-MVS) was developed.

Purpose of the Study:

• To create an organ-level 3D phenotyping workflow for narrow, overlapping leaves.
• To enable accurate, non-destructive measurements for saffron breeding programs.
• To provide a cost-effective and reproducible solution for high-throughput plant phenotyping.

Main Methods:

• Utilized a dual-camera rotary acquisition system for reduced occlusion.
• Employed open-source SfM-MVS reconstruction for 3D model generation.
• Developed a multi-constraint clustering strategy for organ-level segmentation and automated measurements.

Main Results:

• Dual-perspective imaging reduced occlusion errors by 75% compared to single-view.
• Automated leaf measurements showed high agreement with manual references (R² > 0.94, MAPE < 6%).
• Above-ground biomass and canopy volume correlated with corm yield, suggesting resource allocation trade-offs.

Conclusions:

• Organ-level 3D phenotyping of narrow-leaved plants is feasible with low-cost hardware and transparent workflows.
• The workflow is adaptable to other narrow-leaved crops like wheat, rice, and onion, democratizing phenotyping.
• Methodological transparency and cost-effectiveness can achieve high measurement precision, advancing plant phenomics and breeding.