A miniaturized phenotyping platform for individual plants using multi-view stereo 3D reconstruction

Area of Science:

• Agricultural Science
• Plant Science
• Biotechnology

Background:

• High-throughput plant phenotyping is crucial for crop improvement and management.
• Existing phenotyping platforms struggle with small plants and complex structures.
• Automated 3D data acquisition is needed for efficient plant analysis.

Purpose of the Study:

• To develop a miniaturized, automated platform (MVS-Pheno V2) for high-throughput phenotyping of low plant shoots.
• To improve upon previous phenotyping platform designs for enhanced data acquisition and processing.
• To provide a solution for accurate shoot architecture analysis in small plant species.

Main Methods:

• Development of the MVS-Pheno V2 platform based on multi-view stereo 3D reconstruction.
• Integration of hardware with rotation, wireless communication, and a data processing system.
• Calibration of point clouds and extraction of key plant phenotypes (height, leaf area, volume, etc.).

Main Results:

• The MVS-Pheno V2 platform demonstrated high accuracy in point cloud calibration (0.585% MAPE).
• Excellent correlation coefficients (R² > 0.969) were achieved for plant height, leaf length, and leaf width.
• Low root mean squared errors (RMSE) confirmed the precision of measurements for key traits.

Conclusions:

• The MVS-Pheno V2 platform offers a robust solution for high-throughput phenotyping of low-lying individual plants.
• This technology is particularly valuable for plant breeding and management studies focusing on shoot architecture.
• The platform facilitates efficient and accurate data collection for small plant species.