A multiband-pass filtering method to suppress sleeper noise in railway subgrade vehicle-mounted GPR data

Vehicle-mounted ground penetrating radar (GPR) is an effective and rapid tool to detect railway subgrade defects. However, the signal to noise ratio (SNR) of the raw GPR data on railway subgrade is much lower than that of other detection sites. The primary noises are caused by radar wave reflections from the rails and steel-concrete sleepers. However, the rail noise is always invariant since the rails are longitudinal relative to the GPR scanning direction. On the contrary, the sleeper noise is regularly variable since the sleepers are transverse to the GPR scanning position. Therefore, it is essential to suppress the sleeper noise. First, we simulate the GPR signal based on the railway subgrade model and get the synthetic GPR data. There are many “honeycomb” banded distributions in the synthetic GPR image. It is the reflection signal from the sleepers. According to the spectral characteristics of the synthetic GPR data, the sleeper noise is easy to be identified. Next, we design a multiband-pass filter (MPF) to process the synthetic GPR data, and the sleeper noise dramatically weakens. Finally, the real GPR data are filtered by the multiband filter; the “honeycomb” band strips are substantially suppressed. With further image processing the subgrade defects can be more accurately identified.