A robotic field scout has showed potential for identifying potato virus Y (PVY), one of the most significant pathogens affecting potato farming.
PVY is transmitted by aphids and can result in losses from 10-70%. Infected potato farms are often deregistered from seed production, resulting in financial hits for producers, and secondary infections known to follow PVY can compound losses.
Challenges with manual and molecular detection
Tackling infections can be a time-consuming and labour-intensive process, with scouting done manually and suitably qualified experts hard to find. Molecular testing is another option for diagnosing the virus in crops but can prove prohibitively expensive when testing many plants.
To address these issues, a team of researchers in Canada developed the AgriScout robot, which features cameras and a what is known as a real time kinematics or RTK module mounted on four wheels with agriculturally-treaded tyres.
Using AI to identify infected plants
Together, the aim is that the machine automatically moves through fields, providing image capture and geolocation. These images are then sent to a cloud-based server for analysis using a You Only Look Once (YOLO) deep learning model, producing an infestation map of fields which can then be used for targeted disease control measures.
A problem with visual machine learning technologies in agricultural settings to date has been the influence of variable sunlight and weather conditions — particularly pertinent in the case of PVY infection, which causes mottled effects on leaves. The team attempted to counter this with a folding canopy and LED lighting array.
Field trial results and detection accuracy
In field trials, the system had a precision score of 85%, reflecting the number of true positive diagnoses from its predicted positives. Its precision, reflecting the number of true positives from all actual positives, was 0.8 (from a maximum of 1).
“This outcome demonstrates that the network can detect a high proportion of infected plants (limiting false negatives) without inundating the user with excessive false alarms,” the scientists wrote in the journal Computers and Electronics in Agriculture.
From the findings, they suggested that the technology represented a “significant advancement in agricultural automation.”
“The proposed system shifts disease management from labour-intensive processes to cost-effective, automated solutions that minimize human intervention, while maximizing detection accuracy. The effective performance under real field conditions validates its practical utility for precision agriculture,” they added.
Limitations and next steps for deployment
Alongside the positive claims were some concessions about work still needed. The researchers noted a limitation in the technology’s ability to detect early, less visually obvious signs of disease, albeit suggesting this was something that can be improved. It is also dependent on Starlink satellite connectivity, which may limit where it could be used in world. Forthcoming work will see the team train the model on greater numbers of potato cultivars and seek to improve how it can be used in agricultural workflows.
Key takeaways
- AgriScout robot detects potato virus Y using AI and field imaging.
- System achieved 85% precision in real-world field trials.
- Technology automates labour-intensive crop disease scouting.
- Detection accuracy remains lower for early-stage infections.
- Satellite connectivity and model training limit current scalability.
If you’re interested in the impact of robotics on farming, you might also like our stories on robot weeders and how automation could help a Mediterranean labour crisis.
Want to read more stories like this? Sign up to our newsletter for bi-weekly updates on sustainable farming and agtech innovation.
