How can images collected in the field help farmers and advisors decide where to look, what to investigate and how to plan their work?
This is the question at the heart of a trial being presented as part of the OdinsAEye project.
The trial follows the Mariesminde case, covering approximately 1,100 hectares across 90 fields. It demonstrates how camera technology, artificial intelligence, agronomic knowledge and practical farming experience can be brought together to create a better overview of what is happening in the field.
Connecting technology with agriculture
OdinsAEye brings together researchers, technology specialists, agronomists, companies and farmers. The different partners contribute expertise in areas such as cameras and image collection, AI and image analysis, machine integration, weed mapping and practical agricultural operations.
In the Mariesminde case, the project builds on existing mapping from RoboWeedMaps and collaboration with Patriotisk Selskab. The aim is not simply to collect more images, but to turn those images into information that can support decisions.
What should we look at first?
With around 90 fields to manage, one of the first questions is surprisingly simple:
Where should we look first?
Images collected during field operations can provide observations linked to a location and time. This creates a digital record of what has been seen in the field and allows farmers and advisors to investigate areas that require attention.
The demonstration shows how this information can be explored through maps and images, moving from an overview of weed observations to individual fields and specific plants.
Looking closer at the plants
One of the central examples focuses on cornflowers in field 406-0.
Rather than simply presenting the AI’s result, the demonstration shows the images behind the classification — including both clear and more uncertain examples. This makes it possible to examine what the model is actually seeing and gives agronomists an opportunity to challenge and qualify the computer’s assessment.
Images taken before and after treatment can then raise new questions. If cornflowers are found after treatment, for example, the images provide a starting point for investigating factors such as population, timing, application and field conditions.
The value of the images is therefore not that they answer every question automatically, but that they make the questions more precise.
From individual plants to field planning
The trial also demonstrates how information from individual observations can be used at a larger scale.
Different parts of the same field can have very different conditions. For example, weed pressure around a headland may differ from the centre of the field. Concentrated findings can help identify where a closer inspection may be useful.
The next step is to turn this information into a practical work plan.
If several fields have compatible treatment needs and timing, they may be candidates for being handled together. In the demonstration, fields 38-2, 38-3, 38-4 and 38-5 are used as an example of how several fields could potentially share a first treatment while still having different requirements later in the season.
From data to decisions
This is ultimately what OdinsAEye is exploring: how can field data become useful in everyday agricultural decision-making?
The technology can help identify where to look, provide images to support an assessment, and create a more detailed overview of variation within and between fields.
But the farmer and advisor remain an essential part of the process. Their agronomic knowledge is needed to interpret the observations and decide what action, if any, should follow.
The next step is to test this workflow with selected customers in pilot projects, focusing on concrete decisions and measuring how the technology affects the work in the field.
OdinsAEye is working towards a future where the field itself provides more of the information needed to decide what to do next — from individual plants to the planning of an entire working day.
Video in Danish.