RISUD Research Seminar: Using advanced GPR modelling with machine learning for improving infrastructure sensing outcomes

Ground Penetrating Radar (GPR) is one of the most important geophysical methods and its applications are broad and diverse. They range from classic near surface geophysics problems covering areas such as the natural environment, geotechnics, and archaeology, to urban geophysics, security, and non-destructive testing, to even specialised applications in extreme environments, and subsurface planetary exploration of Mars and the Moon with GPR systems onboard exploring rovers. However, infrastructure sensing applications of GPR are mainly the ones that drive most of the related technologies forward as they play a key part to the maintenance and sustainability of our aging build environment.

GPR data interpretation, however, remains a challenging task for many applications, especially where quantitative information about the geophysical targets is required. The role of realistic GPR simulations is becoming pivotal in advancing the GPR interpretation approaches, for both academic research and practical uses. Increasing computational power is making modelling capability much more efficient and accessible, which is leading to the development of new algorithms, better methodological understanding, and education, as well in development of new approaches, new system concepts, and survey modalities. It has particularly benefitted machine learning based approaches which have begun to offer some interesting and exciting possibilities for certain applications.

This talk will provide a short overview of and some recent advances in GPR modelling and present some results and ideas on how GPR modelling can be used to develop new approaches in understanding and improving GPR data interpretation aiming to extract from them more quantitative information.