Related Experiment Video
Updated: Apr 24, 2026
09:21
A Novel Dual-Modal Deep Learning Approach for Real-Time Removal of Hepatic Fluorescence in Indocyanine Green-Guided Laparoscopic Cholecystectomy
Published on: April 17, 2026
24
An environment-guided visual-temporal deep learning framework for early disease detection in greenhouse horticultural
Jiayu Xiang1, Jinrui Ge1,2, Ceteng Fu1
1China Agricultural University, Beijing, China.
Frontiers in Plant Science
|April 23, 2026
Summary
This study introduces a novel framework for early greenhouse crop disease detection by integrating visual data with environmental factors. The approach significantly improves early warning capabilities, enabling timely interventions and reducing pesticide use.
Area of Science:
- Agricultural Science
- Computer Vision
- Environmental Monitoring
Background:
- Early disease detection in greenhouses is crucial for crop yield and quality.
- Subtle visual symptoms and environmental variations challenge conventional detection methods.
- Reducing chemical pesticide input requires precise and timely disease intervention.
Purpose of the Study:
- To develop a visual-environment joint early disease perception framework for greenhouse crops.
- To enhance sensitivity to subtle disease signals by integrating environmental variables.
- To provide a reliable early warning system for greenhouse disease management.
Main Methods:
- An environment-guided visual attention mechanism was employed.
- A spatial-temporal joint modeling strategy was utilized.
- Environmental variables (temperature, humidity, CO2) were integrated as active priors for visual feature learning.
Main Results:
- The proposed framework achieved 91.3% accuracy, 88.9% recall, and 89.8% F1-score in overall disease detection.
- Early-stage detection showed 88.5% precision and 86.1% recall, extending the lead time by 2.7 days.
- Performance significantly outperformed baseline CNN and Transformer models.
Conclusions:
- The integrated visual-environment framework offers a practical solution for early intelligent warning of greenhouse crop diseases.
- The approach enhances detection sensitivity and robustness in complex greenhouse environments.
- This technology holds strong potential for practical deployment in protected horticulture.
