Real time sampling pulsed GPR on drone

The original article is published on the LinkedIn.

The technology of the pulsed GPR is known more than 40 years and its history is the way from analog to digital domain near the ground surface. Many efforts were spent to take off the Ground Penetrating Radar from the ground surface to the air. There are a lot of obstacles on this way. The main one is relatively high frequency of the GPR signal, that force on the antenna to be as close as possible to the ground surface to eliminate the transmitted energy reflection from surface to air. Another one is an air radio noise and as higher GPR lifts up from the ground as higher noises level it gets. The low-altitude flight is able to help, but it is a dangerous task for the manned aircraft. Evolution of unmanned aerial vehicles (drones) brought great possibility to welcome the geophysical payload on board. Both techniques help to acquire the GPR data in a fast and safe manner.

We tried to start from the low frequencies and obviously, expected to get useful results with such an antenna. The first useful data we have acquired with the GPR mounted on the drone DJI M600 Pro. This was a stroboscopic bi-static device with unshielded antenna and changeable dipoles 150 and 75 MHz. It approved the capability of the GPR use on drone, but some features, like its weight 4.5 kg and limitations of the drone manufacturer, designated the need of a lighter and smaller GPR.

The main things that are influencing on the GPR antenna dimensions and weight are its length, which has to correspond to its frequency and antenna configuration – is it bi-static or monostatic one. The traditional bi-static configuration contains both transmitting and receiving antennas, but monostatic device is capable to work with only one antenna, where antenna is used alternately for transmitting of a sounding signal and receiving of the reflected back signals. The final choice of the monostatic configuration for the drone low frequency GPR based on its smaller sizes. 

The main goal of the pulsed GPR is to generate a transmitting pulse and receive a reflected signal back. The reflected signals are traveling from the receiving antenna dipoles to the user’s output device. This way was fully analog in the beginning, but the point of analog to digital conversion slowly moved from the user to antenna side. The longest period of this exciting history is occupied by the stroboscopic sampling technique. The stroboscopic technique needs to trigger transmitter on each sample of the trace to digitize it. It is slow and expensive in time process, because the device spends a lot of efforts to get only one sample. This technique is still in force in nowadays, but a new alternative already commercially approved. It is a Real Time Sampling. It helps to digitize analog signal right on the antenna side, without need to use any other receiving electronics. Its main approach is to connect analog to digital converter right to the dipoles of a receiving antenna and extremely increase clocking speed of the ADC converter. It gives a lot of opportunities to expand dynamic range and speed up the GPR moving speed. The high digital signals traffic is converting to the high stacking (up to the hundreds, thousands and tens of thousands times), that gives a jump for a signal to noise ratio.

Let’s compare the data acquired by both techniques in the monostatic configuration. These data acquired on the same test site without the drone help, close to the ground surface on the altitude of 20 cm. The same environment conditions were registered during the data acquisition. A horizontal high pass filter is applied for both data sets. Gains are aligned and equal as well. As it visible, the left data has not enough dynamic range in comparison with the right one, gain increasing gives signals clipping and high frequency noises only. But the right profile has enough range still to increase the gain and take a look on the deeper signals. The amount of the air noises increases according to the antenna lifting. It directed the development as the monostatic real time sampling GPR with use on the drone. The changeable dipoles principle allows the antenna frequency changing between the flights without device dismounting from the drone.

We have prepared the test site for this task. It is sandy area with low vegetation and buried different objects: the sewage 50 mm air filled pipe, the sewage 50 mm water filled pipe and the metallic wall profile.

The drone flight was right in the center of the objects. Additionally real time sampling monostatic 500 MHz GPR were used to acquire the data in the contact with ground to get precise and detailed profile of the site as the reference. This system helped to visualize all three objects and the structure of the profile section from the sandy test site. All three objects are detected by the 500 MHz antenna and marked on the Fig. 6.a. The air-filled sewage pipe “a” has the weakest signal, as expected due to the small diameter of the pipe. Next one is multi-hyperbola reflection which is more characterized for water-filled non-metallic pipes and coincides with the water filled plastic pipe on the sandy test site. The strong reflection with the inverted phase belongs to the metallic wall profile.

The drone profiles acquired on the same 50 cm altitude at frequencies 300 and 150 MHz. The drone based data needs to be processed due to the monostatic configuration, which gives ringing after switching to receiving mode, as a remain of the antenna excitation by the transmitting pulse. A horizontal high pass filter is applied for both data sets. As the result it shows the detection of two objects only: water-filled sewage pipe and metallic profile. Air-filled pipe is not visible due to its small diameter and weak reflected signal. The visible interfaces are coinciding with the 500 MHz data on the both airborne profiles. The resolution drops down together with the central frequency of the antennas.

At the end, as conclusion, I would like to introduce one more data set that we got on the peat test site where we have magnetometer data acquired from drone as well. As you can see this site is the result of the excavations of the upper layers of the peat in the previous century, when peat processing plant worked in this area near the Riga city.

The data combination was done with the Geolitix software help. It is giving us the good results, where we could focus on two hyperbolas which coincide with the maximums on the magnetometer (Sensys Magdrone R3) gridded data marked as “d” and “e”. These objects are not excavated yet, but such coinciding shows us possibility to use Real Time Sampling GPRs in airborne performance even on such complicated places as peat areas.