Railway ballast fouling detection using GPR data: Introducing a combined time-frequency and discrete wavelet techniques

Ground-penetrating radar has been recently used for quantifying railway ballast fouling conditions. This paper discusses two approaches for processing the railway ballast ground-penetrating radar signal: the discrete wavelet transform method and the short-time Fourier transform method. Ground-penetrating radar field data collected in Wyoming in 2007 by two 2-GHz air-coupled antennas were used to verify both approaches. The collected ground-penetrating radar signals were processed by discrete wavelet transform. The signal standard deviation values were fitted by linear regression of the fouling indices, which were calculated using ground-truth sieving data. The spectrogram generated by short-time Fourier transform was used to analyse the ballast fouling condition level by comparing the energy attenuation speed. It was observed that both discrete wavelet transform and short-time Fourier transform approaches can be used to analyse ballast fouling conditions. The short-time Fourier transform method is easier to visualize, whereas the discrete wavelet transform approach can more quantitatively predict the fouling level. However, the discrete wavelet transform method provides the fouling condition of the whole ballast layer, whereas the spectrogram provides the fouling condition with respect to the depth profile.