Accurate imaging of multicomponent GPR data based on exact radiation patterns

Scalar migration algorithms developed for three-dimensional seismic data are commonly used for imaging ground-penetrating radar (GPR) data. Yet, radar is a vector phenomenon, such that the GPR amplitudes and phases depend on the antenna orientations and wave propagation paths. To address this issue, vector imaging algorithms that fully account for the vector characteristics of GPR data are required. All previously developed vector imaging algorithms are based on far-field approximations of radiation patterns. We demonstrate the limited applicability of these algorithms and introduce a computationally efficient practically exact-field method that accounts for the far-, intermediate-, and near-field contributions to the radiation patterns. To compute rapidly the required "exact" radiation patterns, inverse fast Fourier transforms are applied to their horizontal wavenumber-frequency domain formulations, balancing the tradeoff between accuracy and efficiency by an appropriate oversampling in the wavenumber-frequency domain, and taking advantage of vertical wavenumber phase shifting. We include the exact radiation patterns in a multicomponent vector imaging scheme that jointly images copolarized and cross-polarized data. This scheme is tested on synthetic and field data containing dipping planar and near-planar structures. For both suites of data, high-quality multicomponent images with reflection amplitudes that are nearly independent of the antenna and reflector orientations are obtained.