Interpreting complex, three-dimensional, near-surface GPR surveys: an integrated modelling and inversion approach

With the increasing computational power of modern personal computers, sophisticated modelling and inversion techniques are becoming popular tools for the interpretation of high-resolution, fully three-dimensional GPR surveys. In this paper, we present the latest results of ongoing practical research into the development of novel, integrated, finite-difference time-domain (FDTD) numerical modelling and linear tomographic inversion methods for the interpretation and analysis of near–surface, 3D GPR data. The proposed approach utilizes the Born approximation solution to the inverse-scattering problem and a truncated singular value decomposition (TSVD) to create the final, inverted reconstructions. A three-dimensional, full-field, O(2,4) accurate FDTD modelling scheme is used to generate the numerical-based Green’s functions and incident fields for the inversion. As such, accurate antenna sources (including the influence of shields) and near-field air/ground interface effects are inherently included in the inversion formulation. The performance of this integrated method is evaluated via a simulated, 3D, forensic-based, test-case example (a 900 MHz survey over a clandestine human burial target) including coherent noise from near-surface clutter. Although the example is simplistic, the results show that the scheme works well, despite some assumptions in the inversion methodology. As such, useful information can be gained on the true form, depth, location and spatial interrelationships of the buried features and, therefore, improved interpretations can be obtained in a three-dimensional context.