Ai-powered blackleg disease detection and mapping system for precision potato farming

Overview

This solution addresses blackleg, a destructive potato disease that cannot be controlled through conventional foliar spraying of agrochemicals and therefore requires physical eradication of infected plants. The technology combines artificial intelligence, deep learning, and geographic information systems (GIS) to detect, classify, and geolocate diseased potato plants in the field. By guiding growers and robotic systems to the exact location of infected plants, it enables precise eradication, reduces unnecessary agrochemical use, and helps protect tuber yield. The innovation has clear commercialization potential for the potato industry and broader precision agriculture markets.

Technical specifications

• Deep convolutional neural network (DCNN) models, including GoogLeNet, VGG-16, and EfficientNet, trained using TensorFlow and PyTorch frameworks for image classification
• Image datasets collected from five potato fields across Prince Edward Island and New Brunswick, capturing spatial and temporal variability of healthy plants, blackleg-infected plants, and weeds
• Validated classification accuracy exceeding 90% for distinguishing healthy and diseased potato plants
• Integration with GPS-enabled sensing hardware to geotag infected plants and generate GIS-based disease maps with attribute tables
• Designed to retrofit onto an existing smart sprayer platform and a field buggy equipped with sensors and trained models for real-world evaluation
• Supports both manual eradication guidance and robotic/autonomous eradication workflows

Technology readiness level

The technology has completed preliminary feasibility validation, demonstrating strong DCNN performance in classifying potato plant health under varied field conditions. Current and planned work focuses on extending the models to specifically detect blackleg-infected plants, refining the smart spraying system design to produce GIS-supported geotagged disease maps, and conducting ground-truthing evaluations in laboratory environments. Field-level performance validation using a sensor-equipped buggy is the next milestone toward commercial deployment.