GPR signal de-noising by discrete wavelet transform

• Published in: NDT & E international : independent nondestructive testing and evaluation Vol. 42; no. 8; pp. 696 - 703
• Main Authors: Baili, Jamel, Lahouar, Samer, Hergli, Mounir, Al-Qadi, Imad L., Besbes, Kamel
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.12.2009; Elsevier
• Subjects: Applied sciences; Buildings. Public works; Cross-disciplinary physics: materials science; rheology; DWT; Exact sciences and technology; Fundamental areas of phenomenology (including applications); GPR; Materials science; Materials testing; Measurement and testing methods; Measurements. Technique of testing; Physics; Signal de-noising; Soft threshold; Solid mechanics; Structural and continuum mechanics; Wavelet transform; DWT; GPR; Signal de-noising; Wavelet transform; Soft threshold; Ground penetrating radar; Threshold detection; Pavement; Discrete transformation; Proximity detector; Wavelet transformation; Filter; Bandwidth; Discrete time; Metrology; Signal processing; Buried object; Localization; Root mean square value; Signal analysis; Non destructive test; Structural analysis; Signal to noise ratio
• ISSN: 0963-8695; 1879-1174
• DOI: 10.1016/j.ndteint.2009.06.003

Ground penetrating radar (GPR) is a non-destructive investigation tool used for several applications related to civil infrastructures; including buried objects detection and structural condition evaluation. Although GPR can be effectively used to survey structures, signal analysis can be sometimes challenging. The GPR signals can be easily corrupted by noise because the GPR receiver has usually an ultra-wide bandwidth (UWB). The noise collected by the system can easily mask relatively weak reflections resulting from the inhomogeneities within the surveyed structure; especially when they are at a relatively deep location. This paper presents the use of discrete wavelet transform (DWT) to de-noise the GPR signals. Various mother wavelets were used in this study to de-noise experimental GPR signals collected from flexible pavements. The performance of wavelet de-noising was evaluated by computing the signal-to-noise ratio (SNR) and the normalized root-mean-square error (NRMSE) after de-noising. The study found that wavelet de-noising approach outperforms traditional frequency filters such as the elliptic filter. At the same level of decomposition, the Daubechies order 6 and Symlet order 6 outperform the Haar and Biorthogonal mother wavelets when de-noising GPR signals by soft thresholding.