Harmonising Cross-System GPR Wave Amplitude for Concrete Diagnosis with Machine Learning

Benchmarking the amplitudes of ground penetrating radar (GPR) signals has received little attention in the past, likely because different research groups have used their own GPR data acquired for different projects individually. To facilitate the integration of GPR data for machine learning and change detection through building a functional fingerprinting database, the amplitude attributes must be harmonized across different systems. This would make GPR more useful as an evaluation tool for material states in civil engineering applications. A simple and straightforward method using a metal plate reflector is proposed to convert the amplitudes between different instruments. The method was successful in concrete diagnosis with machine learning. Future work will consider more variables, such as concrete with different rebar diameters at different depths, and fuse multi-frequency data to expedite machine learning.