Related Experiment Video
Updated: Jan 12, 2026
07:36
Visualizing Early Infection Sites of Rice Blast Disease Magnaporthe oryzae on Barley Hordeum vulgare Using a Basic Microscope and a Smartphone
Published on: March 17, 2023
2.2K
Hyperspectral image analysis for classification of multiple infections in wheat
Manon Chossegros1,2, Amelia Hubbard3, Megan Burt3
1Department of Chemical Engineering and Biotechnology, University of Cambridge, West Cambridge Site, Philippa Fawcett Drive, Cambridge, CB3 0AS, UK.
Plant Methods
|November 8, 2025
Summary
Early detection of multiple wheat plant diseases using hyperspectral imaging and deep learning shows promise. EfficientNet models achieved 81% accuracy, aiding early disease recognition for crop yield protection.
Area of Science:
- Agricultural Science
- Computer Vision
- Plant Pathology
Background:
- Plant diseases cause significant economic losses in arable crops.
- Early and accurate disease identification is crucial for effective crop management.
- Distinguishing multiple concurrent infections is challenging but essential for targeted treatments.
Purpose of the Study:
- To investigate the efficacy of hyperspectral imaging and deep learning for classifying multiple concurrent wheat diseases.
- To develop and evaluate deep learning models for identifying single and mixed infections of yellow rust, mildew, and Septoria.
- To explore the impact of co-infections on hyperspectral signatures of wheat pathogens.
Main Methods:
- A dataset of 1447 hyperspectral images of wheat leaves with single and mixed infections was created.
- Four deep learning models (Inception and EfficientNet with 2D/3D convolutions) were trained on the dataset.
- Model performance was evaluated based on overall classification accuracy and accuracy for specific disease combinations.
Main Results:
- EfficientNet with a 2D convolutional input achieved the highest overall classification accuracy of 81%.
- The model demonstrated 72% accuracy in detecting a combined infection of yellow rust and mildew.
- Pathogen hyperspectral signatures were found to be influenced by the presence of other co-infecting pathogens.
Conclusions:
- Hyperspectral imaging combined with deep learning is a viable approach for classifying multiple wheat diseases.
- The developed models show potential for early disease recognition in large-scale farming, even with limited data.
- Further research with larger, balanced datasets is needed to validate findings under field conditions and explore pathogen interactions.
