Content uploaded by Markus Eckerstorfer
Author content
All content in this area was uploaded by Markus Eckerstorfer on Oct 03, 2018
Content may be subject to copyright.
DRONE-MOUNTED UWB RADAR SYSTEM FOR MEASURING SNOWPACK
PROPERTIES: TECHNICAL IMPLEMENTATION, SPECIFICATIONS AND INITIAL
RESULTS
Rolf Ole Rydeng Jenssen1,2,∗
, Markus Eckerstorfer2, Svein Ketil Jacobsen1, Rune Storvold2
1UiT The Arctic University of Norway, Depar tment of Physics and Technology, Tromsø, Norway
2Norut Northern Research Institute, Tromsø, Norway
ABSTRACT: Airborne ground penetrating radar systems allow for carrying out snowpack surveys in complex
terrain. Ultra wideband radars operate within the lower part of the microwave band and are suitable for
measurements of snow depth and layering in a time-saving and safe manner. We have developed a complete
radar system based on a commercial UWB radar sensor, custom designed antennas and a single board
acquisition computer in all weighing 4 kg and fitting (without antennas) into a 30 x 25 x 15 cm3box. The
radar is capable of measurements with a frequency range from 0.95 - 6 GHz, giving roughly 5 cm slant range
resolution and an unambiguous range in air of 5.75 m. The radar can be carried by an octocopter drone
with a wingspan of 1.5 m, flying autonomously at an altitude of 1 m above the snow surface. In this paper
we present the characteristics and specification of our drone-borne radar system and show results from two
different campaigns. We were able to resolve snow stratigraphy in great detail in a dry snowpack, identifying
the most prominent layers. Our second example shows the system’s capabilities of detecting a person buried
under 1.5 m of wet snow.
Keywords: UWB radar, Ground penetrating radar, UAV, drone, Snow stratigraphy
1. INTRODUCTION
Ground penetrating radars (GPRs), especially ul-
tra wideband radars (UWB) operating in GHz-bands
have penetration capabilities and range resolu-
tions that enable information extraction of snowpack
structural features (e.g. Marshall et al., 2007). Thus,
such systems provide a practical alternative to tradi-
tional point-scale measurements that are time con-
suming and influenced by the choice of measure-
ment location. However, GPRs are conventionally
deployed on the ground, by dragging an antenna
with direct ground contact or at a small standoff dis-
tance. In complex terrain, such as rough avalanche
debris, an airborne GPR is of significant advantage
(e.g. Yankielun et al., 2004), as it increases acces-
sibility and decreases deployment time.
We have developed a UWB radar system that
is mountable on a remotely piloted aircraft (RPAS)
(Figure 1), commonly referred to as a drone. By
doing so, we solved the problems of 1) construct-
ing a light, compact and portable radar system, with
2) high range resolution and the ability to penetrate
the snowpack from an airborne platform, as well as
3) an autonomously flying drone with high payload
capabilities and engine redundancy.
∗Corresponding author address:
Rolf Ole Rydeng Jenssen,
UiT The Arctic University of Norway, Depar tment of Physics and
Technology, Tromsø, Norway;
email: rolf-ole.r.jenssen@uit.no
Figure 1: UAV-borne radar system. The UWiBaSS is the grey box
mounted beneath the drone, with the transmitting antenna (grey
plate) and both receiving antennas (black sheets) visible.
2. DRONE-BORNE UWB RADAR SYSTEM
2.1. UWB radar
The UWB radar, or ultra wideband snow sounder
(UWiBaSS), is a GPR that we have developed
for drone-mounted surveys of layered snowpacks
over ground or sea ice (Jenssen et al., 2016).
The radar consists of an m:sequence UWB radar
sensor developed by the German company Ilmsens
(https://www.uwb-shop.com/), custom designed
spiral and Vivaldi antennas, and a single board
acquisition computer with processing software.
Besides weight, size and range resolution, unam-
biguous range and incident power impinging the
target were central design parameters. Unam-
biguous range describes the range from which a
transmitted radar pulse can be reflected and re-
ceived before the next pulse is transmitted. Incident
power at target depends on antenna gain, height
above target (snow surface) and radar system
amplification parameters. These properties dictate
how high the drone can fly above the lowest surface
of interest (typically ground), in our case currently
at a maximum of 5.75 m. Additionally, the unam-
bigeous range of the system inherently affects the
measurements speed of the radar system, which
in turn affects the speed the drone can fly above
the snow surface. In the presented cases, the
maximum speed is about 2-3 m/s due to the current
configuration of the radar. However, this speed can
be increased significantly with asynchronous data
acquisition, which has been implemented and is
ready for use in future campaigns. The radar has
a total of three antennas, of which a planar spiral
antenna is the transmitting antenna and two Vivaldi
antennas act as receiving antennas. The Vivaldi
antennas are mounted in 90 degree offset to each
other to provide reflection polarization capabilities of
the target (Figure 1).The described radar properties
are summarized in Table 1.
Table 1: Main characteristics of the UWiBaSS
Characteristics Value
System bandwidth 5.05 GHz (0.95-6)
Range resolution ≈5 cm
Unambiguous range in air 5.75 m
Weight ≈4 kg
m-sequence clock 13.312 GHz
Measurement rate 32 Hz (max 1000 Hz)
Max power consumption ≈12.7 W (Radar ≈9 W)
Field of view 0.35 m diameter
(from 1 m above surface)
2.2. RPA
The drone currently in use to carry the UWiBaSS
is an octocopter. The ‘Kraken’ octocopter can lift a
maximum payload of 11.5 kg. Each of the eight en-
gines has a maximum rated thrust of 8.45 kg using
18 x 6.1 inch propellers. ’Kraken’ uses 6 cell Li-Pol
batteries (currently at 30 Ahr). For navigation and
control, a ‘pixhawk2’ autopilot running ‘arducopter’
is used. A laser rangefinder, mounted on one of
the eight arms accurately measures the distance to
the ground. It is set up with a ‘Here+’ GPS system
this allows for the use of RTK and very accurate
positioning. ‘Kraken’ can be set up with a ‘MBR
144’ radio system to operate a 15 Mbps radiolink.
3. METHODS
3.1. Campaign setup
Preparatory work on site before mission deploy-
ment takes roughly 15 min, including mounting pro-
pellers and batteries on the drone, antennas on the
UWiBass and setup of the ground control station as
well as radio communication to the airport tower.
Currently, the drone can only be flown in visual line-
of-sight mode (VLOS) as the drone does not have a
camera mounted and lacks obstacle detection sen-
sors. VLOS missions, however, can be flown both
manually and autonomously, the latter scheme fol-
lowing a pre-defined flight path.
The UWiBass can be operated via switches
mounted on the outside (radar on/off, radar control
arm/start/stop). Survey data has to be downloaded
after each mission with a WLAN cable and pro-
cessed for a first quick look. The radar system can
also be operated via secure shell (SSH) and Near
Real-time (NRT) data visualization can be achieved
using the MBR radiolink.
3.2. Postprocessing of radar data
An inherent property of antennas is that all spa-
tial components of the incident field at the receiving
antenna are integrated. As a consequence, a sin-
gle measurement illuminates a 3D volume of snow,
about 0.35 m wide and as deep as the snowpack
is above ground, when the radar is 1 m above the
snow surface. However, only a 1D average of the
returned energy is imaged.
During postprocessing of data, the radar traces are
stacked together to form a 2D image of the snow-
pack. Each pixel intensity is represented in terms of
voltage returned to the antennas. By squaring each
pixel, they are presented in terms of power, which
helps in analysing the data, as some noise is sup-
pressed from the image. For radar images with low
signal-to-noise-ratio (SNR), often due to wet snow,
thresholding further suppresses low level pixels and
thus reduces noise further, while histogram equal-
ization evenly distributes the pixel intensities in the
image to amplify weak returned signals in the snow-
pack.
4. RESULTS
We present data from two different campaigns.
Campaign 1 shows our system’s capabilities to re-
solve snow stratigraphy while campaign 2 focuses
on the detection of a metal object and a buried per-
son within the snowpack.
4.1. Campaign 1 - Snow stratigraphy
Campaign 1 took place in central Svalbard in May
2018 during dry snow conditions. The radar was
mounted on a snowmobile sledge, 50 cm above
snow surface to simulate airborne data acqusition.
The snowpack had variable depths up to a max-
imum of 155 cm. It consisted of wind-deposited
snow at the surface and a gradually more coarse-
grained snowpack at depth with a thick layer of
depth hoar above ground. Three relatively hard lay-
ers (P - K) characterized an otherwise rather soft
snowpack. The radar intensity image (Figure 2)
Figure 2: Radar image showing intensity variations in backscat-
tered energy (yellow means more energy) through snow depth
(y-axis) and distance (x-axis). In-situ measured snow hardness
from a pit dug in the transect is superimposed.
shows high back scattered energy from the snow
surface (air-snow interface) as well as from the un-
dulating snow-ground interface. The three relatively
harder snow layers, as well as the bottom part of the
snowpack, reflected more energy back to the radar
then the softer middle part.
4.2. Campaign 2 - Buried person
Campaign 2 took place on the island of Andøya
in Northern Norway in April 2018 during wet snow
conditions. A person was buried under 1.5 m in a
road embankment, together with a metal plate at
1 m depth (Figure 4). With less than 1 m/s, the
drone was flown over both the metal plate and the
buried person. Below the clearly visible snow sur-
face showing a strong reflection, two hyperbolic re-
flections are visible, indicating the metal plate and
the person buried in the snow (Figure 3). During
this campaign, the snow had up to 8 % liquid water
content and therefore a thresholding procedure was
used to improve visualization of the targets.
Figure 3: Radar image showing variations in backscattered en-
ergy (yellow means more energy) through snow depth (y-axis)
and time (=distance on the x-axis). The red dashed lines indicate
the target hyperbolas typical for strong point reflectors.
5. DISCUSSION
5.1. (In)Capabilities of the UWiBaSS
The UWiBaSS is optimized to resolve detailed snow
stratigraphy as well as to detect buried objects in a
variety of snow conditions. Thus, high vertical res-
olution has been traded off against high penetration
depth in wet snow conditions, which could be ob-
tained using lower radar frequencies (at the cost of
bandwidth).
Weak snow layers are in the order of 1 cm thick; thus
their detection is very difficult. Nevertheless, distinct
layer differences are detectable. Weak snow lay-
ers are often found adjacent to harder layers or right
above or below ice layers. Thus, detecting distinct
hardness changes or ice layers can be used to infer
the presence of a weak snow layer.
The UWiBaSS, as demonstrated above, is also ca-
pable of penetrating wet snow, with liquid water con-
tent of up to 8 %, like in the case of the buried per-
son. It should be noted that the human target was
only visible in one of four passes across the tran-
sect. This can be explained by the low measure-
ment speed of the radar which needed a very low
flight speed to detect the targets.
A limiting factor is that with a current field of view of
about 0.35 m in diameter, a very tight grid needs to
be flown at a distance of 1 m above the snow sur-
face in order to cover an avalanche debris with a
missing person or car. To overcome this problem,
we are currently developing a radar with an ambigu-
ous range in the air of 42 m. This allows to raise the
field of view to 7.14 m and thereby opening up the
grid without changing the range resolution of 5 cm.
5.2. (In)Capabilities of the ’Kraken’ octocopter
The ‘Kraken’ octocopter has been designed to test
the UwiBaSS under controlled conditions, having
enough redundant power to lift the radar also dur-
ing an engine failure. It has not been optimized for
operational use in regard to flight time or for par-
ticular operational scenarios where real-time sensor
navigation, data processing and visualization are
needed. However, we have developed tools for op-
erational use of drones that are currently used for
other applications such as iceberg tracking. These
tools can easily be adapted to provide NRT visual-
ization and mapping of the UWiBaSS data.
Figure 4: Setup of the object burial test with a metalplate and
a buried person 50 cm and 1.5 m below the snow surface. The
drone is hovering over the buried person.
6. CONCLUSION
We have developed a drone-based UWB radar sys-
tem capable of resolving snow stratigraphy and de-
tecting a buried person in a range of snow condi-
tions. Our system can be deployed within roughly
15 min. For a fully operational system, however, we
are currently developing a UWB radar that can be
flown higher above ground, thereby also flying BV-
LOS missions. We are also currently testing a real-
time radar processing unit and live transmission to
an operator screen.
For any of the described applications, a radar ex-
pert is currently needed to interpret the radar data.
Automatic detection of buried persons could possi-
bly be remedied using an artificial intelligence ap-
proach for on-board machine learning interpretation
of the radar signals followed by automatic flagging
and geotagging of objects.
ACKNOWLEDGEMENT
R.O.R.J acknowledges CIRFA for personal funding.
Campaign 1 was carried out within the SIOS Infra-
Nor project. Campaign 2 was carried out during a
workshop organized and funded by the Norwegian
Public Road Administration.
REFERENCES
Marshall, H.P., Schneebeli, M., Koh, G. (2007) Snowstratigra-
phy measurements with high-frequency FMCWradar: Com-
parison with snow micro-penetrometer. Cold Regions Science
and Technology DOI: 10.1016/j.coldregions.2006.08.008
Yankielun, N., Rosenthal, W., Davis, R. E. (2004)
Alpine snow depth measurements from aerial FMCW
radar Cold Regions Science and Technology DOI:
10.1016/j.coldregions.2004.06.005.
Jenssen, R.O.R., Eckerstorfer, M., Vickers, H., Høgda, K.-A.,
Malnes, E., Jacobsen, S.K. Drone-basedUWB radar to mea-
sure snow layering in avalanche start-ing zones Proceedings
of the International Snow Science Workshop, Breckenridge,
CO
