 |
| Titel |
High-resolution GPR imaging using a nonstandard 2D EEMD technique |
| VerfasserIn |
Chih-Sung Chen, Yih Jeng, Hung-Ming Yu |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250073617
|
|
|
|
|
|
| Zusammenfassung |
| * Corresponding author. Tel.: +886-2-7734-6416; Fax: +886-2-2933-3315
Ground Penetrating Radar (GPR) data are affected by a variety of factors. Linear and
nonlinear data processing methods each have been widely applied to the GPR use in
geophysical and engineering investigations. For complicated data such as the shallow earth
image of urban area, a better result can be achieved by integrating both approaches. In
this study, we introduce a nonstandard 2D EEMD approach, which integrates the
natural logarithm transformed (NLT) ensemble empirical mode decomposition
(EEMD) method with the linear filtering technique to process GPR images. The
NLT converts the data into logarithmic values; therefore, it permits a wide dynamic
range for the recorded GPR data to be presented. The EEMD dyadic filter bank
decomposes the data into multiple components ready for image reconstruction.
Consequently, the NLT EEMD method provides a new way of nonlinear energy
compensating and noise filtering with results having minimal artifacts. However, horizontal
noise in the GPR time-distance section may be enhanced after NLT process in some
cases. To solve the dilemma, we process the data two dimensionally. At first, the
vertical background noise of each GPR trace is removed by using a standard linear
method, the background noise removal algorithm, or simply by performing the sliding
background removal filter. After that, the NLT is applied to the data for examining the
horizontal coherent energy. Next, we employ the EEMD filter bank horizontally at each
time step to remove the horizontal coherent energy. After removing the vertical
background noise and horizontal coherent energy, a vertical EEMD method is then
applied to generate a filter bank of the GPR time-distance section for final image
reconstruction.
Two buried models imitating common shallow earth targets are used to verify the
effectiveness of the proposed scheme. One model is a brick cistern buried in a disturbed site
of poor reflection quality. The other model is a buried two-stack metallic target that the signal
reflected from the lower stack is often masked by the upper one. Compared with the results
obtained by using standard processing methods, the proposed approach significantly
improves the target resolution. A field example is also provided to demonstrate the
competence of the proposed scheme.
Keywords: GPR; 2D EEMD; Logarithmic transform; Horizontal coherent energy |
|
|
|
|
|
|