DeTeC  Demining Technology Center

The object used in this acquisition is a PMN-2 mine buried in the loamy soil section of our sandbox at a depth of about 5 cm. We have selected here a data set that gives quite good and understandable results; we warn you that with other type of (smaller) objects in a perhaps more inhomogeneous soil you will get poorer results.
 

Acquisition statistics This first screen was used on DETEC-1 to show the (poor) statistics of the radio modem link. We have kept it in DETEC-2 to show some internal DSP parameters that are not directly accessible by the PC program. It gives information on the number of traces (A-scan) acquired, the gain currently applied to the displayed C-scans, the resolution of the grid along the X-axis (usually 10 mm, but we have also done some tests at 5 mm) and finally the thickness of a C-scan. The palette used to display the image is also shown.

 

Energy We define the energy in each point of the grid as the sum of the square of each samples of the trace located at that point. We use it to select the traces that are presumably not part of the object and so are a quite good representation of the mean background. In contrary to the C-scan pictures, it use a unipolar auto scaled palette where dark blue is the minimum energy value and red is the maximum.

 

Traces selected This image show in blue the traces that have been chosen as explained above as a representation of the mean background. The red area correspond to the traces of higher energy where objects can potentially be located. The green dots are in an area that has not been sufficiently covered by the scans and so hasn't been used by the selection algorithm.

 

Antenna movement This image was useful for DETEC-1, it show in red the point of the grid covered by the antenna. In the case of DETEC-2 the scan method is good enough to have a good coverage and if no problem occurs during the scan, most of the points must be in red.

 

Coverage Again an image very useful for DETEC-1, it gives an indication of the density of the acquisition. The blue area are were there is less traces acquired. Red points are were more traces have been acquired. In DETEC-2, due to the regular movement, the density is generally quite homogeneous. In this data set there is a red area in the lower part of the image because the scan was not started immediately after having pressed the start button. So a big amount of traces have been acquired at the same position (in this case, the average of all these traces is used).

 

Mean height In DETEC-1 again, as the antenna was carried by hand, it was important to know if it has been maintained has a constant height relative to the soil surface. By looking for the first maximum present in the A-scan it is possible to have a good indication of this height if the soil is flat. In DETEC-2 the antenna is kept mechanically at a constant height, so this screen give an idea of the screen roughness. We see in this picture that the computed height decrease from the bottom of the image (start of the scan) to the top (end of the scan); this comes from a thermal drift of the GPR. This effect disappear after a radar warm-up of about 15 minutes. The presence of an object near the surface can also have an influence on this image

 

Diff height If more than one trace is acquired at a given point of the grid, this image shows the antenna height difference between the lower and the higher scan. This was again very useful in DETEC-1 to evaluate the quality of the scan.

 

Last A-scan This image shows the last acquired A-scan. It is here just to show that there is no problem with the radar or our electronic interface.

 

B-scan with fixed Y This image shows 10 regularly spaced B-scans, with Y constant in each scan. This type of image can be useful in order to manually correct the gain applied to the images. We distinguish here that an object is buried in the center of the area, its first red hyperbola being located between position 110 and 120 (each tick at the right of the image is equal to 20 vertical samples) it will be interesting to look at that depth in the C-scan images displayed latter.

 

B-scan with fixed X Same type of image than the previous one, but this time X is constant in each scan. We see here again the object at the same depth. We can conclude from theses two images than something interesting can be seen in the C-scan in the center-right part of the image at a depth offset beginning at 110.

 

C-scan at offset 10 The next displayed image is the computed C-scan at a depth offset of 10 samples (i.e. 10 x 25 ps). It is computed by taking in each point of the grid the mean of a given number of samples (3 in this case) around the given offset. The image is then interpolated from the original grid resolution (40x40) to the screen resolution (480x480) to get smoother shapes. Negative values are displayed in blue, positive in red and value around 0 in green/cyan (the palette is displayed in the acquisition statistics screen).

 

C-scan at offset 116 The user can then move down in the soil by steps of 1 or 10 samples (i.e. 25 or 250 ps). The C-scans above offset 110 show no object, confirming what we have seen on the B-scans. Then, as predicted, we clearly see between 110 and 120 the red reflection of the object.

 

C-scan at offset 129 If we go deeper we see then the blue (opposite polarity) reflection of the object at the same location in the scan area. This blue shape is surrounded by yellow/orange shapes which are in fact the vanishing extremities of the red hyperbola seen above. You may wonder why we don't display the absolute value of the data in order to always see the object reflection in the same color. We tried it; it works well with quite good images like the one we show here, but in case of noisy images it gives a lot of confusion in the interpretation of the images.      An animation is also available.