We would like to find patterns in the vertical direction of the image, representing one GPR trace. If there is an object, we will get a frequency impulse in that direction. A mine also has a characteristic resonance response. This frequency could be detected by a Fourier Transforms (FT).
We tried taking the FT in the vertical direction of sand3_hr. The result can be seen in Figure 15, which represents the absolute value surface plot. Figure 16 shows the phase or angle diagram. In this case, the absolute value is much more significant, the phase diagram looks almost random. This method shows interesting results, as we have several peaks on the absolute value plot, at approximately the position of the objects, but in our case, we can not see any clear characteristic resonance response of our mine in Sand3. Therefore, the method is not really applicable in our case. If we can improve image acquisition, we might keep this method in mind and try it on better data. There is a Matlab function frequences.m, which computes the frequency of the response of the highest peak in the image. The approximate depth of the object can be derived from the characteristic frequency in the FT.
Figure 15: Absolute value of the Fourier
Transform of sand3_hr
Figure 16: Phase of the Fourier
Transform of sand3_hr