PhenoRob-F: An autonomous ground-based robot for high-throughput phenotyping of field crops

Area of Science:

• Agricultural Robotics
• Plant Science
• Genetics

Background:

• Quantitative trait discovery for crops necessitates large, high-quality phenotypic data.
• High-throughput phenotyping platforms are crucial for acquiring this data.
• Autonomous mobile robots offer advantages in payload capacity, flexibility, and proximity for high-resolution imaging.

Purpose of the Study:

• To introduce PhenoRob-F, a novel autonomous wheeled robot for efficient field phenotyping.
• To demonstrate PhenoRob-F's capability in autonomous operation and data acquisition under diverse field conditions.
• To validate the robot's performance in various crop species and phenotyping tasks.

Main Methods:

• Developed PhenoRob-F, a cross-row robot with integrated visual and satellite navigation for autonomous operation.
• Captured RGB, RGB-D, and near-infrared spectral data from wheat, rice, maize, and rapeseed canopies.
• Utilized machine learning models (YOLOv8m, SegFormer_B0) for image analysis and 3D reconstruction for trait calculation.

Main Results:

• Achieved high precision (0.783), recall (0.822), and mAP (0.853) for wheat ear detection.
• Obtained high mIoU (0.949) and accuracy (0.987) for rice panicle segmentation.
• Calculated plant height with high accuracy (R²=0.99 for maize, R²=0.97 for rapeseed) and classified drought stress in rice with >97% accuracy.

Conclusions:

• PhenoRob-F is a highly effective tool for high-throughput field phenotyping.
• The robot provides precise phenotypic data crucial for trait analysis and crop improvement.
• Autonomous phenotyping robots like PhenoRob-F significantly advance genetic studies and breeding programs.