Mobile Soil Health: Visual and Contextual Intelligence for Wheat Pest and Soil Management

What this project explored

This project investigated the technical feasibility of integrating visual and contextual information using advanced data fusion techniques in a mobile pest management solution. The aim was to provide:

• Quantitative analysis — rapid detection and effective quantification of wheat pests.
• Pest management context — placing observations within regionally and nationally relevant pest tolerance thresholds.
• Decision support — guidance on whether a pesticide application is advised.

Innovations

Lightweight pest quantification. A new optimised lightweight CNN model designed for rapid, accurate wheat pest quantification on mobile devices, suitable for smallholder farms in areas with limited network coverage.

Robust data fusion. A broad-learning data-fusion approach that combines local activity features with contextual information, supporting more accurate and robust pest detection in field conditions.

Sustainable pest management. Methods for forecasting wheat pest tolerance thresholds and estimating the likely effectiveness of pesticide application after detection. Efficient, sustainable crop protection has significant economic and ecological value in global food production.

Project team

The project team comprised the University of Sheffield (UoS), ADAS and Mutus Tech.

• The University of Sheffield’s Department of Computer Science led the pest recognition and data fusion technique design, drawing on the PestNet model developed in an earlier UK–China smart agriculture project.
• ADAS — a leading independent UK agricultural consultancy — contributed pest modelling expertise, including work on stripe rust and wheat bulb fly.
• Mutus Tech contributed machine-learning and data analytics capability.

Partners

Funded by Innovate UK; delivered with the University of Sheffield and RSK ADAS.