Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pheno-Robot: An Auto-Digital Modelling System for In-Situ Phenotyping in the Field

Yaoqiang Pan, Kewei Hu, Tianhao Liu, Chao Chen, Hanwen Kang

TL;DR

Pheno-Robot addresses the challenge of scalable, high-fidelity field phenotyping by integrating environmental perception, motion planning, and in-situ phenotyping modelling into a cohesive autonomous system. It couples a BEV-based 3D-ODN for instance detection with a graph-mapped representation, a global-local trajectory planner that optimizes view-quality under traversability constraints, and a NeRF-based modelling pipeline (enhanced with occlusion regularization and few-shot learning) to reconstruct detailed plant geometry from sparse field views, using Instant-NGP for efficiency. Experimental results in greenhouse-like and field environments demonstrate accurate perception (e.g., $T= extbf{T}$ with $P$-level metrics), robust navigation (replanning at 5 Hz, speeds up to $1$ m/s), and high-quality in-situ phenotype models (PSNR $>$ 23.5 dB, converging within minutes). The approach enables automated, repeated, high-resolution phenotyping in realistic agricultural contexts, with potential to accelerate agro-genetic studies and precision agriculture workflows; future work will add robotic manipulation and further refine the NeRF pipeline to reduce artefacts under challenging canopies.

Abstract

Accurate reconstruction of plant models for phenotyping analysis is critical for optimising sustainable agricultural practices in precision agriculture. Traditional laboratory-based phenotyping, while valuable, falls short of understanding how plants grow under uncontrolled conditions. Robotic technologies offer a promising avenue for large-scale, direct phenotyping in real-world environments. This study explores the deployment of emerging robotics and digital technology in plant phenotyping to improve performance and efficiency. Three critical functional modules: environmental understanding, robotic motion planning, and in-situ phenotyping, are introduced to automate the entire process. Experimental results demonstrate the effectiveness of the system in agricultural environments. The pheno-robot system autonomously collects high-quality data by navigating around plants. In addition, the in-situ modelling model reconstructs high-quality plant models from the data collected by the robot. The developed robotic system shows high efficiency and robustness, demonstrating its potential to advance plant science in real-world agricultural environments.

Pheno-Robot: An Auto-Digital Modelling System for In-Situ Phenotyping in the Field

TL;DR

Pheno-Robot addresses the challenge of scalable, high-fidelity field phenotyping by integrating environmental perception, motion planning, and in-situ phenotyping modelling into a cohesive autonomous system. It couples a BEV-based 3D-ODN for instance detection with a graph-mapped representation, a global-local trajectory planner that optimizes view-quality under traversability constraints, and a NeRF-based modelling pipeline (enhanced with occlusion regularization and few-shot learning) to reconstruct detailed plant geometry from sparse field views, using Instant-NGP for efficiency. Experimental results in greenhouse-like and field environments demonstrate accurate perception (e.g., with -level metrics), robust navigation (replanning at 5 Hz, speeds up to m/s), and high-quality in-situ phenotype models (PSNR 23.5 dB, converging within minutes). The approach enables automated, repeated, high-resolution phenotyping in realistic agricultural contexts, with potential to accelerate agro-genetic studies and precision agriculture workflows; future work will add robotic manipulation and further refine the NeRF pipeline to reduce artefacts under challenging canopies.

Abstract

Accurate reconstruction of plant models for phenotyping analysis is critical for optimising sustainable agricultural practices in precision agriculture. Traditional laboratory-based phenotyping, while valuable, falls short of understanding how plants grow under uncontrolled conditions. Robotic technologies offer a promising avenue for large-scale, direct phenotyping in real-world environments. This study explores the deployment of emerging robotics and digital technology in plant phenotyping to improve performance and efficiency. Three critical functional modules: environmental understanding, robotic motion planning, and in-situ phenotyping, are introduced to automate the entire process. Experimental results demonstrate the effectiveness of the system in agricultural environments. The pheno-robot system autonomously collects high-quality data by navigating around plants. In addition, the in-situ modelling model reconstructs high-quality plant models from the data collected by the robot. The developed robotic system shows high efficiency and robustness, demonstrating its potential to advance plant science in real-world agricultural environments.
Paper Structure (26 sections, 19 equations, 11 figures, 1 table, 2 algorithms)

This paper contains 26 sections, 19 equations, 11 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Robotic Automation in Agriculture
  4. In-situ phenotyping
  5. Methods
  6. Environment Perception
  7. 3D Detection Network
  8. Graph Mapping
  9. Trajectory Planning
  10. Global Path Planning
  11. Local Trajectory Generation
  12. Feasibility checking
  13. In-situ Phenotyping Model
  14. Neural Rendering
  15. Few-shot learning from Sparse Views
  16. ...and 11 more sections

Figures (11)

  • Figure 1: Pheno-Robot system operates in the greenhouse.
  • Figure 2: System Overview of the Pheno-Robot system.
  • Figure 3: Illustration of EPM principle.
  • Figure 4: Illustration of MPM principle.
  • Figure 5: Illustration of terrain analysis, a) the points of robot surroundings, b) the risk area is converted into polygons.
  • ...and 6 more figures