Satellite imagery recognition project for DAERA

DAERA is looking for innovative solutions to prevent livestock disease outbreaks and to protect the environment. We are tackling this challenge by creating a neural network Proof of Concept that could analyse satellite images.

author photo
Written by Objectivity
Added
The project's goal

DAERA looking for innovative solutions

The Department of Agriculture, Environment and Rural Affairs (DAERA) is responsible for food, farming, environmental, fisheries, forestry and sustainability policy and the development of the rural sector in Northern Ireland. The Department assists the sustainable development of the agri-food, environmental, fishing and forestry sectors of the Northern Ireland economy, having regard for the needs of the consumers, the protection of human, animal and plant health, the welfare of animals and the conservation and enhancement of the environment.

To fulfil its remit, DAERA must be aware of the location and numbers of grazing cattle and sheep. Counting livestock, recording their location and detecting movement of livestock has, until now, been a labour-intensive activity which has required manual counting in the fields. This method is expensive and prone to error.

DAERA and Objectivity Limited have partnered to apply innovation to solve this problem. Using a combination of satellite imagery and artificial intelligence, the solution can automatically detect grazing cattle and sheep, record where the animals are located and monitor livestock movement on a map-based system. The solution should increase efficiency and reduce costs for DAERA by:

  • Monitoring large geographic areas using automated artificial intelligence and accurately detecting livestock without human intervention.
  • Reducing the need for physical inspections and manual livestock counting to validate subsidy claims.
  • Improving the assessment and management of disease outbreaks, enabling authorities to target their actions.
  • Monitoring ammonia emissions to provide environmental controls.

We are currently prototyping phase 1 of the entire project. We are training models for the object detection and semantic segmentation with the state-of-the-art convolutional architectures including YOLO, ResNet and UNET. The aim is to create a neural system that, given an image, recognises and locates the presence of the livestock from satellite imagery. After detecting livestock on a satellite image, the number of animals will be counted and this information stored, along with the corresponding coordinates. This solution will enable DAERA to have an accurate assessment of the livestock numbers, locations and movements.

After we confirm the feasibility of our solution, we hope to move into phase two in which we will extend the solution to detect habitat change and further support DAERA’s mission.

The project is a part of the Small Business Research Initiative that enables the public sector to leverage innovative technology and speed up its adoption.

How can we help you?

Nigel Lomas
Public Sector Lead
Top