Area Statistics Deep Learning (ADELE)

Project

Summary

The FSO’s land use statistics are an invaluable tool for long-term land observation. This project involves learning and mastering the use of artificial intelligence (AI) technologies to eventually automate (even partially) the visual interpretation of aerial images in order to detect and classify changes.

Description

Among the learning methods that have emerged from AI for image recognition, Deep Learning is particularly adapted to statistics on land use and cover for which learning data are available in very large quantities. This project precedes the development of an externally commissioned IT prototype that is expected to demonstrate this approach for the national territory as a whole at the start of 2019.

The following steps are planned as part of the project:

  • familiarisation with aerial images (primary data) and the results of previous surveys (reference data);
  • development of a dataset that will be introduced into modelling including data transformation and cleaning;
  • testing modelling techniques and varying their parameters to explore possible solutions;
  • taking delivery of the IT prototype in 2019 and applying the acquired knowledge to verify its robustness.

Objectives

The objective of the project is to introduce staff from the GEO section to new AI technologies with a view to integrating these into production during 2019.

The following objectives should be fulfilled:

  • understanding of the job, issues and success criteria in the form of a problem to be solved in terms of data and methods originating from AI;
  • preparing a partial dataset and set of tools (framework) for use in training, analysis and exploration.