A Vision-Based Automatic Landing Method For Fixed Wing UAVs

J Intell Robot Syst (2010) 57:217231
DOI 10.1007/s10846-009-9382-2
A Vision-Based Automatic Landing Method
for Fixed-Wing UAVs
Sungsik Huh David Hyunchul Shim
Received: 1 February 2009 / Accepted: 1 August 2009 / Published online: 23 October 2009
Springer Science + Business Media B.V. 2009
Abstract In this paper, a vision-based landing system for small-size xed-wing
unmanned aerial vehicles (UAVs) is presented. Since a single GPS without a
differential correction typically provide position accuracy of at most a few meters, an
airplane equipped with a single GPS only is not guaranteed to land at a designated
location with a sufcient accuracy. Therefore, a visual servoing algorithmis proposed
to improve the accuracy of landing. In this scheme, the airplane is controlled to y
into the visual marker by directly feeding back the pitch and yaw deviation angles
sensed by the forward-looking camera during the terminal landing phase. The visual
marker is a monotone hemispherical airbag, which serves as the arresting device
while providing a strong and yet passive visual cue for the vision system. The airbag is
detected by using color- and moment-based target detection methods. The proposed
idea was tested in a series of experiments using a blended wing-body airplane and
proven to be viable for landing of small xed-wing UAVs.
Keywords Landing dome Fixed-wing UAVs Visual servoing Automatic landing
1 Introduction
It is well known that the landing is the most accident-prone stage for both manned
and unmanned airplanes since it is a delicate process of dissipating the large amount
of kinetic and potential energy of the airplane in the presence of various dynamic and
operational constraints. Therefore, commercial airliners heavily rely on instrument
landing system (ILS) wherever available. As for UAVs, a safe and sound retrieval of
S. Huh (B) D. H. Shim
Department of Aerospace Engineering, KAIST, Daejeon, South Korea
e-mail: hs2@kaist.ac.kr
URL: http://unmanned.kaist.ac.kr
D. H. Shim
e-mail: hcshim@kaist.ac.kr
Reprinted from the journal 217
the airframe is still a signicant concern. Typically, UAVs land manually by either
internal or external human pilots on a conventional runway or arresting mechanisms.
For manual landing, the pilot obtains visual cue by naked eye or through the live
images taken by the onboard camera. Piloting outside of the vehicle needs a lot
of practice due to the limited situation awareness. As a consequence, a signicant
portion of mishaps happen during the landing phase. Nearly 50% of xed-wing
UAVs such as Hunter and Pioneer operated by the US military suffer accidents
during landing. As for Pioneer UAVs, almost 70% of mishaps occur during landing
due to human factors [1, 2]. Therefore, it has been desired to automate the landing
process of UAVs in order to reduce the number of accidents while alleviating the
pilots load.
There are a number of automatic landing systems currently in service. Global
Hawk relies on a high-precision differential GPS during take-off and landing [3].
Sierra Nevada Corporations UCARS or TALS
1
are externally located aiding sys-
tems consisting of a tracking radar and communication radios. It has been success-
fully used for landing of many military xed-wing and helicopter UAVs such as
Hunter or Fire Scout on runways or even on a ship deck. The information including
the relative position or altitude of the inbound aircraft is measured by the ground
tracking radar and relayed back for automated landing. Some UAVs are retrieved
in a conned space using a special net, which arrests the vehicle own manually
by the external pilot. Scan Eagle is retrieved by a special arresting cable attached
to the tall boom, to which the vehicle is precisely guided by a differential GPS.
2
These external aids listed above rely on special equipment and radio communication,
which are not always available or applicable due to complexity, cost, or limits from
operating environment. Therefore, automatic landing systems that are inexpensive,
passive, and reliable are highly desired.
Vision-based landing has been found attractive since it is passive and does not
require any special equipment other than a camera and a vision processing unit
onboard. A vision-enabled landing system will detects the runway or other visual
markers and guide the vehicle to the touchdown point. There are a range of
previous works, theoretical or experimental, for xed-wing and helicopter UAVs
[46]. Notably, Barber et al. [7] proposed a vision-based landing for small xed-wing
UAVs, where visual marker is used to generate the roll and pitch commands to the
ight controller.
In this paper, a BWB-shaped xed-wing UAV is controlled to land automatically
by using vision algorithms and relatively simple landing aids. The proposed method
is shown to be viable without resorting to expensive sensors and external devices.
The overall approach, its component technologies, and the landing aids including the
landing dome and recovery net are described in Section 2. The vision algorithms
including detection and visual servoing algorithm which enables airplane to land
automatically are described in Section 3. Finally, the experiment results of vision-
based landing test are shown to validate the algorithm in Section 4. In Section 5,
conclusion and closing remarks are given.
1
http://www.sncorp.com
2
http://www.insitu.com/scaneagle
218 Reprinted from the journal
2 System Description
The proposed landing system consists of three major components: an inated dome
as a visual marker, a vision processing unit, and a ight controller using a visual
servoing algorithm. In order to develop a low-cost landing system without expensive
sensors or special supporting systems, the vision systemis integrated with an MEMS-
based inertial measurement unit (IMU) and a single GPS. Without a differential
correction, the positioning accuracy of GPS is known to be a few meters horizontally
and twice as bad vertically. Therefore, even if the position of the dome is accurately
known, it will be still difcult to hit the dome consistently with the navigation system
that has position error larger than the size of the landing dome.
For landing, with roughly estimated location of the dome, the airplane starts ying
along a glide slope leading to the estimated position of the dome. When the vision
system locks on the dome, the ight control switches from glide slope tracking to
direct visual servoing, where the offset of the dome from the center of the image
taken from the onboard front-looking camera is used as the error signals for the
pitch and yaw control loops. In the following, detailed descriptions on the proposed
approach are presented (Fig. 1).
2.1 Airframe and Avionics
The platform used in this research is a BWB-based UAV (Fig. 2). BWBs are
known to carry a substantially large payload per airframe weight due to the extra
lift generated by the airfoil-shaped fuselage. Our BWB testbed is constructed with
sturdy expanded polypropylene (EPP), which is reinforced with carbon ber spars
embedded at strategic locations. The vehicle is resilient to shock and crash, a highly
welcomed trait for the landing experiments in this paper.
This BWB has only two control surfaces at the trailing edge of the wing, known
as elevons, which serve as both aileron and elevator. At the wingtips, winglets are
installed pointing down to improve the directional stability. The vehicle has a DC
brushless motor mounted at the trailing edge of the airframe, powered by Lithium-
polymer cells. The same battery powers the avionics to lower the overall weight. The
Initial
Estimation
Beginning of
glide slope
Terminal visual
servoing
Air dome
Fig. 1 Proposed dome-assisted landing procedure of a UAV
Fig. 2 The KAIST BWB (blended wing-body) UAV testbed
large payload compartment (30 15 7.5 cm) in the middle of the fuselage houses
the ight computer, IMU, battery, and radio receiver. These parts are all off the shelf
products.
The avionics of KAIST BWBUAVconsists of a ight control computer (FCC), an
inertial measurement unit(IMU), a GPS, and a servo controller module. Flight con-
trol computer is constructed using PC104-compatible boards due to their industry-
grade reliability, compactness and expandability. The navigation system is built
around a GPS-aided INS. U-blox Supersense 5 GPS is used for its excellent track-
ing capability even when the signal quality is low. Inertial Sciences MEMS-IMU
showed a robust performance during the ight. The vehicle communicates with the
ground station through a high-bandwidth telemetry link at 900 MHz to reduce the
communication latency (Table 1, Fig. 3).
The onboard vision system consists of a small forward-looking camera mounted
at the nose of the airframe, a video overlay board, and a 2.4 GHz analog video
transmitter with a dipole antenna mounted at a wingtip. Due to the payload limit,
the image processing is currently done at the ground station, which also serves as
the monitoring and command station. Therefore the image processing and vision
algorithm results are transmitted back to the airplane over the telemetry link. Image
processing and vision algorithms are processed on a laptop computer with Pentium
Core2 CPU 2 GHz processor and 2 GB memory. Vision algorithm is written in MS
Visual C++ using OpenCV library.
3
It takes 240 ms to process the vision algorithm
at an image, then, the visual servoing command is transmitted back to the vehicle at
4 Hz, which is shown to be sufcient for the visual servoing problem considered in
this research.
3
Intel Open Source Computer Vision Library
Table 1 Specication of BWB
testbed
a
http://www.haakworks.com
Base platform StingRay 60 by Haak Works
a
Blended wing-body (BWB)
Dimensions Wing span: 1.52 m(W)
Wing area: 0.52 m
2
Aspect ratio: 4.43
Weight 2.6 kg (fully instrumented)
Powerplant Axi 2217/16 DC brushless motor
12 6 APC propeller
Hacker speed controller X-40-SB-Pro
Lithium-ion-polymer (3 S: 11.1 V 5,350 mAh)
Operation time Over 10 min
Avionics Navigation: single GPS-aided INS
GPS: U-Blox Supersense 5
IMU: inertial science MEMS IMU
Differential/absolute pressure gauges
Flight computer: PC104 Pentium-M 400 MHz
Communication: 900 MHz Ethernet Bridge
Vision system Color CCD camera (70 deg FOV)
2.4 GHz analog transmitter
Frame grabber: USB 2.0 analog video
capture kit
Operation Catapult-launch, body landing
Autonomy Speed, altitude, heading, altitude
hold/command
Waypoint navigation
Automatic landing
2.2 Landing Equipment
In order to land an airplane in a conned space, an arresting device is needed to
absorb the kinetic energy in a relatively short time and distance. In this research, we
propose a dome-shaped airbag, which offers many advantages for automatic landing
scenarios.
The dome is constructed with sturdy nylon of a vivid red color, which provides a
strong visual cue even at a long distance. It also absorbs the shock during the landing
and therefore the vehicle would not need to make any special maneuver but simply
y into the dome at a low speed with a reasonable incident angle. The vision system
runs a fast and reliable color tracking algorithm, suitable for high-rate visual servoing
under various light conditions.
For operation, the bag is inated within a few minutes by a portable electric
blower or possibly by a gas-generating charge and transported to anywhere in a
compact package after deated. Using the hooks sewn around the bottom patch,
it can be secured to the ground using pegs. Its dome shape allow landing from any
direction without reinstallation while conventional arresting nets have to be installed
to face the wind to avoid crosswind landing. Since this approach is vision-based, the
landing dome is not equipped with any instruments, signal sources, or communication
devices. Therefore it does not consume any power or emits energy that can be
detected by adversaries.
Fig. 3 Avionics and hardware architecture
In addition, a recovery net can be used to capture larger vehicles. In this case,
the landing dome serves only as a visual cue and the recovery net is responsible for
arresting the vehicle.
3 Vision Algorithms
The proposed vision algorithm consists of a color-based detection, a moment-based
detection, and visual servoing. The color- and moment-based methods perform in
a complementary manner to detect the landing dome more reliably. In the terminal
stage of landing, the vehicle maneuvers by the visual servoing command only (Fig. 4).
3.1 Color-Based Target Detection
The domes color is the strongest visual cue that distinguishes the target from other
objects in the background. In RGB coding, a pixel at image coordinate (x, y) has
three integers, (I
R
, I
G
, I
B
), varying from 0 to 255, respectively. As the pixels that
belong to the red dome vary quite large depending on lighting condition, shadow,
color balance, or noise, a ltering rule is needed to determine whether a pixel
Fig. 4 Proposed landing dome and a recovery net
belongs to the dome or not. Based on many images of the dome taken under various
conditions, the following ltering rule is proposed:
aI
B
(x, y) < I
R
(x, y) 255
b I
G
(x, y) < I
B
(x, y) 255
0 c < I
R
(x, y)
(1)
where (a, b, c) = (1.5, 1.5, 20). The threshold levels are chosen rather inclusively for
detection of wider color range. Figure 5 visualizes the colors that are qualied as a
dome using Eq. 1.
3.2 Moment-Based Target Detection
When multiple red objects are found by the color-based detection described above
in a given image, another classication is needed to determine whether an object is
0
100
200
0
100
200
0
50
100
150
200
250
B
L
U
E
RGB Color Space
RED
GREEN
0
50
100
150
200
250
0
100
200
0
50
100
150
200
250
B
L
U
E
RGB Color Space
RED
GREEN
c
b
a
Fig. 5 Full RGB color space (left) and selected region by ltering rule (right)
Original
image
Thresholded
image
Filled by
morphology
Result
image
Fig. 6 Moment-based image processing procedure
indeed the landing dome or not. For this, an image moment-based ltering is devised.
Among many algorithms such as template matching or image moment comparison,
Hus method [8, 9] is used since it is computationally efcient for high-rate detection
needed for vision-based landing (Table 1).
Hus moment consists of seven moment invariants derived from the rst three
central moments. In a gray image with pixel intensities I(x, y), raw image moments
M
ij
are calculated by
M
ij
=
y
x
i
y
j
I (x, y). (2)
Then simple image properties derived by raw moments include the area of image
M
00
, and its centroid { x, y} =
M
10
M
00
, M
01
M
00
.
Central moments for digital image are dened as below
ij
=
x x
y y
q
f (x, y) (3)
Table 2 Hus moment value
of various shapes in Fig. 6
Roof Road sign Automobile Landing dome
Hus moment 1 0.48560 0.20463 0.27423 0.21429
Hus moment 2 0.18327 0.00071 0.04572 0.01806
Hus moment 3 0.01701 0.00009 0.00193 0.00106
Hus moment 4 0.00911 0.00028 0.00098 0.00007
Fig. 7 Application of Hus method to domes seen from various viewpoints
and the central moments of order up to the third order are given as
01
=
10
= 0
11
= M
11
xM
01
= M
11
yM
10
20
= M
20
xM
10
02
= M
02
yM
01
21
= M
21
2 xM
11
yM
20
+2 x
2
M
01
12
= M
12
2 yM
11
xM
02
+2 y
2
M
10
30
= M
30
3 xM
20
+2 x
2
M
10
03
= M
03
3 yM
02
+2 y
2
M
01
(4)
Fig. 8 Autopilot loops for longitudinal (top) and lateral dynamics (bottom) of airplane
In order to make the moments invariant to both translation and changes in scale,
the following formula is used to divide the corresponding central moment by the
properly scaled moment.
ij
=

ij
(
(i+j+2)
/
2
)
00
(5)
Since the dome is projected to be somewhere between a semicircle to a full circle, its
Hus moments would not vary too much regardless of the distance and the viewing
angle. Therefore Hus moment can be a good invariant characterization of a target
shape. Using the second- and third-order moments given in Eq. 5, the following Hus
image moments that are invariant to translation, rotation, and scale are derived:
1
=
20
+
02
2
=

20
02
2
+
2
11
3
=

30
3
12
2
+
3
21
03
4
=

30
12
2
+
21
+
03
2
(6)
In this paper, only four image moments are used for detection.
Hus moments approach is applied to various red objects seen fromair as shown in
Fig. 6 and their moment values are listed in Table 2. The rooftop and the automobile
in Fig. 6 has relatively larger Hus moments while the road sign and the landing
dome have comparable rst moments due to their round shapes. The higher-order
moments of the road sign has much smaller values than those of landing dome since
the road sign has a smaller area for given number of pixels. Although the landing
dome seen directly above would look much similar to the road sign, such cases are
excluded because the camera is mounted to see only the side of the dome during
landing.
In Fig. 7, the image processing results of landing domes seen from various view-
points. Once a database of Hus moments of these domes was built, the average,
standard deviation, and minimum/maximum values of landing dome can be deter-
mined. Based on these values, the blob with the largest area in a given image is nally
declared as the landing dome. In case of all candidates are not in the proper range
of Hus moments, the algorithm concludes that a landing dome does not exist in
that image.
im
Y
C
Z
im
X
T
x

f

T
y
C
Y
C
X

,
ref
ref
Fig. 9 Visual servoing scheme
0 5 10 15
10
20
30
40
50
60
70
80
time [sec]
a
l
t
i
t
u
d
e

[
m
]

alt.: Kalman filter
alt.: GPS
alt.: Ref. trajectory
landing at
t=13.2 sec
automatic
landing stage
vision-based
landing stage
Fig. 10 Experiment result of vision-based landing: vehicle altitude
3.3 Visual Servoing
Visual servoing is a method to control a robot using computer vision. The robot is
controlled by the error of features between the desired image and current image. The
visual servoing method can be applied to the xed-wing UAV such that the vehicle
is controlled by the pitch and roll angle errors estimated from an image as shown in
Fig. 8.
As illustrated in Fig. 9, the offset of the landing dome fromthe center of the image
represents the heading and pitch deviation.
= arctan
y
T
= arctan
x
T
(7)
0
50
100
-250
-200
-150
-100
-50
0
40
60
80
LCC x
LCC y [m]
L
C
C

a
l
t

[
m
]
Fig. 11 Experiment result of vision-based landing: 3D trajectory
0 5 10 15
-50
0
50
100
R
o
l
l

[
d
e
g
]

Ref.
A/C
0 5 10 15
-100
-50
0
50
P
i
t
c
h

[
d
e
g
]

Ref.
A/C
0 5 10 15
-200
0
200
Y
a
w

[
d
e
g
]
time [sec]

Ref.
A/C
Fig. 12 Experiment result of vision-based landing: vehicle attitude and heading
0 5 10 15
-20
0
20
b
o
d
y

v
e
l

[
m
/
s
]

v
x
v
y
v
z
0 5 10 15
-50
0
50
b
o
d
y

a
c
c

[
m
/
s
]

a
x
a
y
a
z
0 5 10 15
-500
0
500
a
n
g
u
l
a
r

r
a
t
e

[
r
a
d
/
s
]
time [sec]

p
q
r
Fig. 13 Experiment result: body velocity, body acceleration and angular rate
When the heading and pitch angle errors are zero, i.e. = = 0 (the landing
dome appears at the center of the image), the vehicle ies directly to the landing
dome.
Fig. 14 Sequences of images during landing to the dome (left) and the recovery net (right)
If the camera is mounted at the nose of the airplane so that the Z
C
-axis in Fig. 9
aligns with the direction of ight, any heading and pitch angle deviations indicate
that the heading of the vehicle deviates from the target point. Therefore, in order to
guide the vehicle to the target point, the controllers should minimize these angles. It
can be done simply by sending the heading and pitch angles to the heading and pitch
angle controllers. Since the pitch and yaw deviation angles computed from an image
are directly sent to the controllers, the INS output is not utilized at all. This approach
is simple and robust to any failure in the INS or GPS. This scheme is surprisingly
simple but highly effective as will be shown in the experimental result.
4 Vision-Based Landing Experiment Result
Using the experiment setup described in Section 2, the proposed vision-based landing
algorithm with landing aids was tested. First, the inated landing dome was installed
in the test eld. After the GPS is locked on and the INS is initialized, the vehicle is
launched using a catapult at an airspeed excess of 20 m/s. Once airborne, the vehicle
is commanded to look for the landing dome in its view. When the dome is located
in an image, it generates a glide slope from its current location to roughly estimated
location of the dome. During descent, the aircraft ies by visual servoing, and nally,
it lands at the dome or at the recovery net safely.
In Fig. 10, vehicle altitude as the results of ight test is shown. Before t = 6.6 s,
the vehicle follows the glide slope leading to the roughly estimated location of the
landing dome. During t = 6.613.2, the vehicle ies in the direct visual servoing
mode. At t = 13.2 s, the vehicle contacts with the landing net safely.
The reference altitude and the actual one are plotted in Fig. 10. The altitude of
the landing dome was initially estimated to be 45 maccording to reference trajectory,
but it turned out to be 50 m according to the altitude of Kalman lter at the time of
impact. This observation advocates our approach to use direct visual servoing instead
of following blindly the initial trajectory, which does contain errors large enough to
miss the landing net. Figures 11, 12, and 13 show the vehicles 3D trajectory and
states. In Fig. 14, the images taken from the onboard cameras are shown and the
green rectangles in onboard images showthat the landing dome was reliably detected
during ight. The result also shows that the landing method using both landing dome
and recovery net are feasible to recover the airplane safely.
5 Conclusion
In this paper, a vision-based automatic landing method for xed-wing unmanned
aerial vehicle is presented. The domes distinctive color and shape provide unmis-
takably strong and yet passive visual cue for the onboard vision system. The vision
algorithm detects the landing dome reliably from onboard images using color- and
moment-based detection method. Since the onboard navigation sensors based on
low-cost sensors cannot provide accurate enough navigation solutions, a direct visual
servoing is implemented for precision guidance to the dome and the recovery net.
The proposed idea has been validated in a series of experiments using a BWB
airplane to show that the proposed system is viable approach for landing.
Acknowledgement The authors gratefully acknowledge for the nancial support by Korea Ministry
of Knowledge and Economy.
References
1. Williams, K.W.: A summary of unmanned aircraft accident/incident data: human factors implica-
tions. U. S. Department of Transportation Report, No. DOT/FAA/AM04/24 (2004)
2. Manning, S.D., Rash, C.E., LeDuc, P.A., Noback R.K., McKeon, J.: The role of human causal
factors in U. S. army unmanned aerial vehicle accidents. USAARL Report No. 200411 (2004)
3. Loegering, G.: Landing dispersion resultsglobal Hawk auto-land system. In: AIAAs 1st Tech-
nical Conference and Workshop on Unmanned Aerial Vehicles (2002)
4. Saripalli, S., Montgomery, J.F., Sukhatme, G.S.: Vision-based autonomous landing of an un-
manned aerial vehicle. In: Proceedings of the IEEE International Conference on Robotics &
Automation (2002)
5. Bourquardez, O., Chaumette, F.: Visual servoing of an airplane for auto-landing. In: Proceedings
of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems (2007)
6. Trisiripisal, P., Parks, M.R., Abbott, A.L., Liu, T., Fleming, G.A.: Stereo analysis for vision-based
guidance and control of aircraft landing. In: 44th AIAA Aerospace Science Meeting and Exhibit
(2006)
7. Barber, B., McLain, T., Edwards, B.: Vision-based landing of xed-wing miniature air vehicles. In:
AIAA Infotech@Aerospace Conference and Exhibit (2007)
8. Hu, M.-K.: Visual pattern recognition by moment invariants. In: IRETransactions on Information
Theory (1962)
9. Flusser, J.: Moment invariants in image analysis. In: Proceedings of World Academy of Science,
Engineering, and Technology (2006)

A Vision-Based Automatic Landing Method For Fixed Wing UAVs

Uploaded by

Copyright:

Available Formats

A Vision-Based Automatic Landing Method For Fixed Wing UAVs

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Vision-Based Automatic Landing Method For Fixed Wing UAVs

Uploaded by

Copyright:

Available Formats

J Intell Robot Syst (2010) 57:217231

You might also like