Icon AgrML

Practical example and current challenges

  • Government subsidies from the autonomous province of Bolzano are an important incentive for farms located in South Tyrol. Some provincial subsidies vary according to the size of the cultivated area and the type of farming. The correct allocation of public subsidies to South Tyrolean farmers is therefore an important issue that must be closely monitored to ensure efficient management of public funds
  • One objective of the project AgriML - Machine-Learning for Agriculture in South Tyrol is to extend the automatic recognition of the cultivation forms to 100% declared on the territory of the Autonomous Province of Bolzano. Another objective is to automate the control of the cultivation forms to verify both their correctness and any subsequent changes to them.
  • At present, agricultural subsidies are paid according to a manual procedure based on the guidelines of the European Commission. In this process, periodic overflights are conducted (an area-wide update of South Tyrol takes place approximately every 3 years) to verify farmers' self-declarations regarding the extent and type of crops present in their fields. The overflight images are manually evaluated and labeled. On-site inspections are carried out.
  • A proof of concept (PoC) was developed to demonstrate the feasibility of adopting machine learning techniques in recognizing the type of crops on South Tyrol's satellite imagery and to support and automate the process of granting subsidies to farmers.
  • In the implementation phase, anonymized crop polygons were used to retrieve satellite images of the corresponding cultivated fields on the territory of the Autonomous Province of Bolzano. At the same time, convolutional models and recurrent neural networks were implemented and trained to model the relationship between the satellite images and the crop types present in the corresponding polygon labels. When selecting the infrastructure and architecture of the machine learning solution, it is necessary to ensure that these are appropriate for both data pre-processing and virtual machines for training the algorithms.
Infographic: AgriML

What added value does the "Gaia-X project" offer?

  • Based on Gaia-X, the use case can realize the recognition of crop types in satellite imagery and thus lead to savings in the rental and use of aircraft as well as to a better quality of work for public service employees. Another immediate benefit would be a considerable saving of work for the provincial administration and more satisfaction for farmers
  • The added value of Gaia-X is the possibility to implement a solution based on European sovereignty and absolute data protection. As a European infrastructure, Gaia-X can offer data lake functionalities and scalable computing capacity for training artificial intelligence algorithms.
  • The Gaia-X infrastructure is open to both open source and commercial software and complies with European principles and laws: This use case helps to create a standardization basis for the process of providing agricultural contributions according to the European Commission directives, but fits into the local context of each public administration.
  • The "AgriML" service described in this use case, realized on the basis of Gaia-X, will be available and usable for any public administration of the European Union. In addition to the agricultural sector of South Tyrol, this solution can potentially be applied and add value in the field of green area management, forestry, and the monitoring of the geological characteristics of the territory.

Use Case Team

  • Stefan Gasslitter – Südtiroler Informatik AG – Informatica Alto Adige SpA