Improved GPR image focussing with repetitive normalised Superimposition techniques

Abstract

The Superimposition technique offers an alternative to Fast Fourier Transform (FFT) and Inverse FFT calculations. side lobe levels are reduced with little or no increase in main lobe width, as opposed to standard windowing techniques where side lobe level reduction produces an increase in main lobe width with a resulting loss in resolution. The new technique uses repetitive superimpositions showing improvements over spatially variant apodization (SVA) techniques. A new normalisation scheme enhances side lobe reduction even further with no increase in main lobe width indeed it can reduce the main lobe width. The technique is seen to be more resilient to noise when appropriate multiple evaluations are chosen. The technique produces responses from reflections in GPR data that are resolved to responses much closer to a delta function than FFT/IFFT or SVA evaluations. When used in focussing algorithms the traditional hyperbolic characteristics of a B-scan are focussed into responses whose width in depth and plan position that are slightly better than half a wavelength of the bandwidth used. This is seen in theoretical data and in both data measured by commercial GPRs and in experimental data from a step frequency continuous wave based GPR. Theoretically the technique produces a strong indication of the permittivity of the ground the GPR measures are taken over, while in measured data the identification of the permittivity of the ground is less clear.