Deep Learning for interpretation of subsurface radar data

Purpose and goal

The aim of the project was to develop and use AI and Deep Learning techniques for automatic interpretation of groundradar data, the goal being to locate and characterize buried infrastructure. The outcome of the project has been very positive, and as a direct result of the work, Guideline Geo has presented a unique technology for locating cables and pipes underground using AI-based interpretation of radar data. The technology has been protected under the MALÅ AI brand and will be an integral part of the company´s products in the near future.

Expected results and effects

In the project we have developed new AI-based techniques for interpreting ground radar data. We have further implemented these in a software library and a reference implementation, MALÅ AI Sandbox, which we use for continuous improvement of the AI core, as well as demonstrations for end-users.

Planned approach and implementation

The project has run as planned with some minor deviations from the planned work. Based on initial very promising results we have reduced the scope of the work on synthetic data. We have also seen that unsupervised training right now is not as effective as supervised, which is why we focused on the latter in our reference implementation.