Removal of ringing noise in GPR data by signal processing

摘要

Ringing is a common type of coherent noise in ground penetrating radar (GPR) data. When this kind of coherent noise is strong and is not properly removed, deeper structure may be completely masked. Ringing appears as nearly horizontal and periodic events, which are the most important features enabling us to remove the noise by signal processing. In this study, we have reviewed basic principles of various signal processing techniques to remove the ringing noise and compared their performances using field GPR data contaminated by severe ringing noise. The reviewed methods include background removal. f-k filtering, predictive deconvolution with filtering in wavenumber domain, and filtering by radon transform. Furthermore, it is shown that ringing can be successfully removed by the eigenimage filtering method, where GPR image is decomposed into eigenimages by singular value decomposition. This comparative analysis shows that the refined techniques are definitely more effective than the simple methods for the ringing noise removal with less distortion of GPR signals and each method has its own advantage as well as limitations. Moreover, preservation of the horizontally linear events from geological targets can be possible only through a kind of selective or local filtering such as the eigenimage filtering method.