Related Experiment Video
Updated: Jun 11, 2025
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
9.2K
GSP-AI: An AI-Powered Platform for Identifying Key Growth Stages and the Vegetative-to-Reproductive Transition in
Liyan Shen1, Guohui Ding1, Robert Jackson2
1College of Engineering, College of Agriculture, Academy for Advanced Interdisciplinary Studies, Plant Phenomics Research Centre, Nanjing Agricultural University, Nanjing 210095, China.
Plant Phenomics (Washington, D.C.)
|October 10, 2024
Summary
A new AI model accurately predicts wheat growth stages and flowering time using drone imagery and climate data. This tool aids breeders and growers in timely crop management and selection for improved wheat production.
Area of Science:
- Agricultural Science
- Plant Breeding
- Artificial Intelligence in Agriculture
Background:
- Wheat (Triticum aestivum) is a crucial global staple crop requiring precise growth monitoring for management and breeding.
- Accurate, scalable assessment of wheat growth stages (GSs) at the plot level is essential for research and crop improvement.
- Current methods for monitoring wheat development can be labor-intensive and challenging across diverse environments and seasons.
Purpose of the Study:
- To develop a multimodal deep learning model (GSP-AI) for identifying key wheat growth stages and predicting flowering time.
- To create a comprehensive Wheat Growth Stage Prediction (WGSP) dataset for training and validating the model.
- To provide a reliable and scalable toolkit for plot-level wheat growth stage assessment.
Main Methods:
- Established the WGSP dataset with 70,410 annotated images from wheat varieties across China, UK, and US, coupled with climate data.
- Developed a deep learning architecture using Res2Net and Long Short-Term Memory (LSTM) to integrate visual and climatic data.
- Trained and validated the GSP-AI model on multiseasonal data (2018-2021) and further refined it using smartphone imagery.
Main Results:
- The GSP-AI model achieved 91.2% accuracy in identifying key GSs and predicted flowering days with a 5.6-day RMSE.
- Refinement with smartphone images improved accuracy to 93.4% for GS identification and reduced flowering prediction RMSE to 4.7 days.
- The model effectively learned from canopy images and climatic patterns for robust growth stage prediction.
Conclusions:
- The GSP-AI model offers a significant advancement for breeders and growers in monitoring wheat development.
- This tool facilitates informed agronomic decisions and effective crop selection by providing timely growth phase information.
- The study demonstrates the potential of multimodal deep learning for precise, plot-level crop management and wheat improvement.
