Ground-penetrating radar (GPR) is a common technique for evaluating the structure and quality of civil engineering materials. The ever-increasing demand for higher GPR time resolution and better interpretation of GPR data has motivated the use of advanced signal processing methods for GPR applications. In this article, we review the major advances in signal processing techniques employed in civil engineering for different tasks, such as estimation of thickness, permittivity, and roughness. Their performance is tested and compared through numerical testing and using experimental data from laboratory measurements.

In the field of civil engineering, sounding the pavement layers is classically performed using standard ground-penetrating radar, whose vertical resolution is bandwidth dependent. The layer thicknesses are deduced from both the time delays of backscattered echoes and the permittivity of layers. In contrast with conventional spectral analysis approaches, this letter focuses on one of the machine learning algorithms, namely, the support vector machine, to perform time delay estimation and dielectric constant estimation of the medium from backscattered radar signals. This letter shows the super time resolution capability of such technique to resolve overlapping and fully correlated echoes within the context of thin pavement layer testing.

In this letter, we propose a new efficient method to estimate the direction-of-arrival (DOA) and range of near-field sources in a decoupled way. First, a non-Hermitian cumulant matrix is constructed, whose eigenvectors associated with zero eigenvalues are used to directly estimate the DOA by using the MUSIC algorithm. Then, the ranges are estimated with the estimated DOA by orthogonalizing the remained eigenvectors. Compared with other modified 2-D MUSIC, the proposed algorithm can greatly reduce the computational complexity by avoiding a 2-D search with only one matrix and one eigenvalue decomposition. Simulation results show the effectiveness of the proposed method.

In the electromagnetic field, radar is widely used to measure or estimate the media parameters or to detect targets through obstructions. For horizontally stratified media, the layer thickness can be deduced from the time delays of backscattered echoes and the dielectric constants. The high-resolution method estimation of signal parameters via rotation invariance techniques (ESPRIT) has been proposed for time-delay estimation. In practice with a radar, backscattered echoes are correlated. In order to apply the ESPRIT method, in this letter, we propose to use two adaptive improved spatial smoothing techniques with the propagator method for fighting against the correlation between the echoes. The proposed solution does not use any approximation. Numerical examples are provided to show the performance of the algorithm.

In civil engineering, conventional methods used to estimate the thickness of pavements assume flat interfaces. In contrast, this study uses a rigorous electromagnetic method called propagation-inside-layer-expansion (PILE) to simulate the radar backscattered signal at nadir from a rough pavement made up of two rough interfaces separating homogeneous media. The statistical distribution of the first two echoes is studied by comparison with the default flat case, together with their frequency behaviour. Within the scope of road pavement survey by ground penetrating radar, the scattering model is finally used to assess the performance of the estimation of signal parameters via rotational invariance techniques (ESPRIT) algorithm, one of the well-known high-resolution time-delay estimation techniques.