Affordable smartphone-based 3D phenotyping tool for grape bunch architecture analysis

Overview

This solution offers an affordable 3D phenotyping tool that reconstructs the three-dimensional architecture of fruit bunches from standard 2D images captured by a smartphone or non-professional camera. Designed for table and wine grape growers, the tool addresses a critical need in precision viticulture: quantifying bunch shape and compactness to support better crop management decisions.

By providing accurate 3D reconstructions at low cost, the technology enables growers to optimize berry spatial distribution within bunches. This improves ventilation and light penetration, reduces disease vulnerability, lowers chemical inputs, and ultimately enhances fruit quality and consumer appeal. The approach replaces expensive laboratory equipment such as micro-CT scanners with a field-deployable alternative accessible to commercial growers.

Technical specifications

Core capabilities:

• Constructs 3D point clouds of grape bunches from multi-angle sequential 2D images
• Achieves point cloud accuracy of less than 1mm
• Generates between 500,000 and 800,000 data points per bunch, capturing fine structural detail
• Tested across 10 typical bunch shapes representing a cross progeny of 160 grapevine genotypes
• Validated against micro-CT reference scans for three distinct grape cultivars
• Algorithms under development for 3D point cloud segmentation and improved reconstruction from 2D inputs

Key advantages:

• Uses widely available consumer cameras or smartphones rather than specialized imaging equipment
• Enables non-invasive phenotyping of bunches over their entire developmental stages
• Supports field-based in situ image capture, removing the need to excise samples
• Provides quantitative compactness metrics superior to traditional visual scoring

Technology readiness level

The technology has completed initial proof-of-concept validation. Researchers have demonstrated 3D bunch architecture reconstruction for three cultivars and tested reconstruction algorithms on 10 representative bunch shapes from a 160-genotype progeny. Current work focuses on scaling validation across the full progeny, transitioning from excised bunches in controlled imaging rooms to in situ field capture, and refining segmentation and reconstruction algorithms. Additional development is needed for customized data acquisition equipment and field-ready workflows. Estimated timeline for full validation is at least one year. The team is actively seeking partnerships for equipment development, algorithm improvement, and key personnel funding.