96 Transportation Research Record 2289 response of EM wave in ballast. The scattering analysis approach showed potential to distinguish fouled ballast from clean ballast. To overcome the limited penetration depth problem of 2-GHz antennae, lower-frequency antennae such as 500-MHz horn antennae can also be used to obtain comprehensive information for the ballast, subballast, and subgrade (2). Al-Qadi et al.(11), Xie et al.(2), and Leng and Al-Qadi (12) used a time–frequency approach to interpret the GPR data. They found that this approach can characterize the signal in time and frequency domains simultaneously and quantify the fouling and moisture content.

In summary, GPR is a great tool for assessing the railroad ballast condition. However, the processing and interpreting of GPR data are challenging. Currently, there are three primary approaches for data interpretation: traditional approach, scattering approach, and time–frequency approach. The traditional approach interprets the data in the time domain. The thickness of the clean ballast can be determined if a clear interface between clean and fouled ballast is observed in the GPR image. The scattering approach interprets GPR data in the time domain and provides fouling depth on the basis of the difference in scattering intensity (10). The time–frequency approach applies short-time Fourier transform (STFT) to track the frequency change with time (2). When fouling is present, energy drops at the fouling location in the STFT image so the fouling depth can be obtained. These three methods can be combined for data interpretation, but the methods have certain limitations. It is not unusual for the gradation of fouled ballast to change gradually, which results in no clear interface between clean and fouled ballast. Therefore, it is very difficult to obtain fouling depth with the traditional approach. When the scattering approach is used, fouling depth is determined at the location where scattering intensity changes. To see the change in scattering intensity, users need to choose the appropriate color map and display parameters of the GPR image. Yet different users may choose different color maps and parameters and thus interpret the information differently. Therefore, the result could be user dependent, especially when scattering is not clear enough. Although the time–frequency approach, which uses the STFT method, can provide more accurate fouling information, it can process only one scan per time (2). The time–frequency approach is suitable for assessing the fouling condition at specific locations, but it can be challenging to analyze the GPR data for an entire railroad network with this technique. The objective of this study is to develop a new data processing technique that overcomes the aforementioned limitations. The wavelet technique, which is one of the most efficient data analysis techniques, is proposed for GPR data interpretation. Controlled laboratory tests were conducted at the Illinois Center for Transportation. Wavelet transform was then applied to the collected data to extract useful information that can indicate fouling conditions. Finally, the proposed wavelet technique was validated with field GPR data collected from the Orin subdivision in Wyoming.