Artificial Intelligence to Reduce the Use of Herbicides

The Danish – and highly innovative – project RoboWeedMaPS wants to develop a system that automatically recognizes and detects weeds in the field using artificial intelligence. New ways and synergies are to be developed in the area of precision farming.

Says Birger Hartmann from Agrinavia ‘we are excited and proud to be part of this cutting edge project with the aim to reduce the use of herbicides in favor of both society, environment and the economy of the individual farmer. The consortium expects a reduction of up to 80% on the use of herbicides at the production site.

Research and Development

R&D is carried out at Aarhus University under the leadership of Senior Scientist Rasmus Nyholm Jørgensen. The ambitious project is partly funded by The Innovation Fund Denmark and extends over a period of four years. The project was launched last year, however builds upon previous work done at Aarhus University. The work within the project includes:

  • Development of a high-tech camera
  • Development of an artificial intelligence model
  • Recognition of up to 100 different weed plants
  • Compilation of weed maps
  • Optimal calculation of control strategy
  • Compilation of application maps
  • Optimization of spray technology with injection.


The development of the artificial intelligence is the core of the project. That means developing a system, that automatically recognizes the weed and the different weed plants that are registered in the field. The method used is called Deep Learning – a neural network, artificial intelligence, is presented to and stores a very big amount of data, Big Data. In this case several thousand photos of different crops and weeds. Thereby the intelligence is trained to recognize or to find new connections based of what it has seen – been presented to. 

The ’camera system’ will recognize different kind of weed and on different stages of growth. This is far more effective than anthropogenic information, since a computer can store so much more information and data and furthermore find cohesions in big amount of data.

 In the Field

The expectation to the end product is a workflow, that makes it possible to photograph and spray in the field in one and the same working process: On the tractor is placed a camera, that will be able to recognize weed plants, when the sprayer drives over them. The software in the sprayer adjusts its treatments while driving. In this way specific weed plants can be targeted exactly where the problem area is located and with herbicides most effective for controlling that specific weed plant. However, during the four-year period of the project, there will be several milestones to pass and component parts to develop. For instance, a weed map that converts to an application map for the sprayer.

The process from weed identification to spraying:

  • Photography of plants: The camera shoots one picture per 10 meters and with 40 kilometers per hour
  • Identification of weed using artificial intelligence
  • Mapping of different weed plants, population and growth stages
  • Calculation of herbicides/pesticides and doses
  • Compilation of application map with multiple herbicide options
  • Precision spraying
  • Documentation of treatment.

Says Birger Hartmann, ‘at Agrinavia we are developing the online solution, that makes it possible for the farmer in one workflow only to carry out the absolutely necessary weed spraying and at the same time have the weed plants and treatment shown on a map. This has been a vision and wish for many years, but not possible because of lack of data.’

Export to foreign markets is also the plan going forward. 

Partners in the project

Besides Aarhus University and Agrinavia, four other companies take part in the project. Each company with their specific knowledge and expertise. The companies are AgroIntelli, IPMConsult, I•GIS and Danfoil.

Weeds Seen from Above – Automatic Recognition with Drone Plane