Effective quality assurance and quality control inspections of new roads as well as assessment of remaining service-life of existing assets is taking priority nowadays. Within this context, use of ground penetrating radar (GPR) is well-established in the field, although standards for a correct management of datasets collected on roads are still missing. This paper reports a signal processing method for data acquired on flexible pavements using GPR. To demonstrate the viability of the method, a dataset collected on a real-life flexible pavement was used for processing purposes. An overview of the use of non-destructive testing (NDT) methods in the field, including GPR, is first given. A multi-stage method is then presented including: (i) raw signal correction; (ii) removal of lower frequency harmonics; (iii) removal of antenna ringing; (iv) signal gain; and (v) band-pass filtering. Use of special processing steps such as vertical resolution enhancement, migration and time-to-depth conversion are finally discussed. Key considerations about the effects of each step are given by way of comparison between processed and unprocessed radargrams. Results have proven the viability of the proposed method and provided recommendations on use of specific processing stages depending on survey requirements and quality of the raw dataset.

No abstract

Ground-penetrating radar ͑GPR͒ sections encounter a resolution reduction with depth because, for electromagnetic ͑EM͒ waves propagating in the subsurface, attenuation is typically more pronounced at higher frequencies. To correct for these effects, we have applied a spectral balancing technique, using the S-transform ͑ST͒. This signal-processing technique avoids the drawbacks of inverse Q* filtering techniques, namely, the need for estimation of the attenuation factor Q* from the GPR section and instability caused by scattering effects that result from methods of dominant frequencydependent estimation of Q*. The method designs and applies a gain in the time-frequency ͑t-f͒ domain and involves the selection of a time-variant bandwidth to reduce high-frequency noise. This method requires a reference amplitude spectrum for spectral shaping. It performs spectral balancing, which works efficiently for GPR data when it is applied in very narrow time windows. Furthermore, we have found that spectral balancing must be applied prior to deconvolution, instead of being an alternative technique.

No abstract

No abstract