In the beginning of the project, Martin Fritzsche of Daimler Benz was interested in a collaboration with the LAMI. He sent us our first GPR created images. Those pictures gave us a good basis to start working on the project. Later, we needed to take our own pictures in order to understand the problem domain better and to perform our own tests.
In collaboration with the EMPA
in
Dübendorf, we took a first set of images. Johannes Hugenschmidt is
the expert at EMPA to take GPR images of road profiles. He uses a
GSSI SIR 10 system with a 900 MHz antenna. Further, EMPA has a large
sand box, 3x2x1.5 meters large, which was ideal for our first
measurements. In Figure 2 is a picture of the GSSI SIR 10
system we used to acquire the images. Figures 3 and 4
show how we moved the antenna in the sand box. The distance was
measured by an incremental encoder connected to the GSSI system,
ensuring that a trace was taken every 1.5 mm. This was necessary to
correct the irregularities due to the manual motion of the antenna. The system
recorded 512 samples in 25 nanoseconds. The resolution of each sample
point was 16 bit. Mr. Hugenschmidt chose these values by
experience. The 25 ns record time for each trace was chosen
empirically on site, to include the response of the floor of the sand
box in the image.
Figure 3: B. Gros moving the antenna, assisted by C. Bruschini
Figure 4: A. Perrig moving the GPR antenna
During our experiences, we tried out many different objects, to get a good understanding of the workings of the GPR. The best image that resulted was sand3, where we put screws, a copper wire, a mine and a rock 8 cm below the surface of the sand. The configuration can be seen in Figure 6. Figure 5 shows the mine we received from the Swiss Army for that day.
The GSSI system uses an image format called RADAN. We wrote PERL
scripts and C programs to transform the test image we received from
EMPA. The other images we received were in the SIR10 DZT File
format. We wrote conversion programs for this file format in C and
Matlab. The Matlab routines sirload256 and sirload16 are
printed in Appendix A. The sirload256 routine
reads an 8 bit SIR10 DZT file into a Matlab matrix
and sirload16 reads the 16 bit file format.
The resulting image, sand3, is displayed in Figure 7. The other images we took did not turn out as well as sand3. There are several reasons why some of the pictures did not turn out well. It was the first time we took our own GPR images. In order to have the GPR sender and receiver as close as possible to the ground surface, we took the EMPA assembled box apart. The incremental encoder was assembled on that box as well, so we needed to hold it by hand (see Figure 4 or 3). Further, the encoder was sliding on the sand, giving uneven signals. In the beginning, we also used small objects, which were difficult to spot later in the images. We also noticed that digging the objects deeper increased the signal quality. All in all, we learned our first lesson in taking GPR images.
We will use Sand3 throughout the rest of this report as sample image. All the images were created by the Matlab code listed in Appendix B.
