Signal fusion using extrapolation with deterministic deconvolution on multi-frequency Qinghai-Tibet Railway GPR data for permafrost subgrade detection

The Qinghai-Tibet Railway, the highest railroad in the world, stretches 1,956 kilometers from Xining to Lhasa in China. Of this ~2000 km total length about one half of it (~960 km) is on permafrost subgrades. Frozen earth disaster and other subgrade diseases lead to subgrade instability and affect the security of railway transportation. It is critical to develop a rapid and efficient technique for railway foundation diseases detection. Multi-frequency ground penetrating radar (GPR) is a good choice to account for both penetration and resolution to image railway subgrades. In general, the lower frequency GPR signal has deeper penetration and the higher frequency GPR signal has higher resolution. We propose a signal fusion method to extrapolate the higher frequency GPR signal to greater depth with a higher resolution that the lower frequency signal provides using the extrapolation with deterministic deconvolution (EDD). In this paper we first introduce the principles of EDD. Second, we discuss the data preparation with pre-processing and filtering to enable the field GPR signal on Qinghai-Tibet Railway being suitable for carrying out EDD. Finally, the multi-frequency radar signal with high signal to noise ratio is fused based on EDD. Both deep penetration and high resolution radar profile is expected. This approach provides a helpful tool for recognizing subgrade disease in different depth under the Qinghai-Tibet Railway.