Location: Crop Production and Protection
Project Number: 0500-00102-001-092-A
Project Type: Cooperative Agreement
Start Date: Aug 10, 2026
End Date: Aug 9, 2027
Objective:
To develop and validate the first AI-based RCR detection and mapping model using multispectral satellite imagery to accurately estimate disease severity and spatial distribution across soybean production environments.
Approach:
The project will be conducted in four integrated phases. First, approximately ten soybean fields with confirmed RCR will be selected across Illinois, with additional sites in Kentucky for environmental contrast, and geo-referenced disease ratings will be collected at 50–60 locations per field during reproductive growth stages. Second, high-resolution PlanetScope multispectral imagery will be acquired, corrected, and used to derive vegetation indices sensitive to canopy stress. Third, a random-forest model will be trained and validated using spatially independent cross-validation to generate pixel-level RCR probability maps and field-level severity estimates, with UAV-derived RGB imagery used as an independent reference. Finally, model performance will be evaluated using accuracy, precision, sensitivity, and F1-score, and outputs will be summarized into field-scale disease maps.