Automatic detection of hyperbolas in georadargrams by slant-stack processing and migration

Ground penetration radar (GPR) is an important tool in geophysical prospecting to examine the structure of the subsurface and to locate objects in the ground at shallow depths. For analysis and interpretation of GPR data, considerable experience is needed. The analysis is time-consuming and expensive for the examination of large areas, preventing more widespread use of georadar. Therefore, automatic data processing and interpretation are the obvious choice for effective application of georadar in civil engineering. The important task in this context is the detection of man-made objects in the ground such as pipes, cables, barrels and ducts. GPR data are commonly composed of bed reflections, diffraction hyperbolas from small objects, and noise. Data processing can be improved by separating these components, and for this purpose slant stack and migration offer important advantages. These algorithms focus specific data components, while simultaneously dispersing the remaining features. To separate data components, invertible transforms are needed. The slant stack focuses linear bed reflections (Gardner & Lu 1991), which enables them to be separated from diffractions and noise. The detection and extraction of the diffractions can be performed by migration, which focuses the hyperbolas. This operation requires a correct estimate of the speed of wave propagation in the ground. This method was proposed by Harlan et al. (1984) for application to seismic data and was also succesfully applied by Meldahl (1993). However, their iterative algorithm requires a lot of computation. For implementation on GPR data, a more practical direct method is needed which could be executed on a personal computer.