High-Precision Phenotyping of Grape Bunch Architecture Using Fast 3D Sensor and Automation

Area of Science:

• Agricultural Science
• Plant Breeding
• Computer Vision

Background:

• Grapevine bunch rot risk is reduced by looser bunch architecture.
• Accurate phenotyping of grapevine bunch traits is crucial for breeding but is labor-intensive and time-consuming.
• Existing phenotyping methods lack the speed and precision required for large-scale plant selection.

Purpose of the Study:

• To develop a fast and high-precision phenotyping pipeline for grapevine bunch architecture.
• To automate the extraction of key 3D bunch traits using 3D scanning and analysis software.
• To validate the developed method for its reliability and applicability in both lab and field conditions.

Main Methods:

• Utilized an Artec Spider 3D scanner to capture dense 3D point clouds of grapevine bunches.
• Developed automated analysis software (3D-Bunch-Tool) to extract 3D bunch traits (berry number, diameter, volume, bunch dimensions).
• Validated the pipeline on diverse grapevine cultivars and breeding material under lab and field conditions.

Main Results:

• Obtained reliable phenotypic data with high significant correlations (up to r² = 0.95 for berry number) compared to ground truth.
• Demonstrated comparable precision between lab-based and direct field applications of the 3D scanning method.
• The developed pipeline enables non-invasive, non-contact phenotyping of large plant sets.

Conclusions:

• The developed 3D phenotyping pipeline offers a fast, high-precision, and reliable method for assessing grapevine bunch architecture.
• This technology facilitates efficient selection of grapevine cultivars with desirable bunch traits, particularly for disease resistance.
• The non-invasive field application significantly enhances the scalability of grapevine breeding programs.