Aerospace and Electronic Systems Magazine September 2016 - 8

Prototype Sense and Avoid System Onboard a ScanEagle Unmanned Aircraft
collision. During the threat aircraft's approach it will remain fixed
at a particular point on the pilot's windscreen. The SAA system
exploits the constant-bearing behaviour of collision-course aircraft
by detecting the relatively stationary features in the camera's field
of view (FOV). Objects that move rapidly across the SAA system's
FOV, however, are not collision threats and, as such, are not detected by the SAA system.
The architecture of the SAA system is based on acquiring precisely synchronised images from the EO camera and aircraft state
data from the GPS/INS system [9], [10]. The integrated greyscale
imagery and state data then undergoes multiple stages of consecutive processing by the single board computer: stabilization, morphological filtering, temporal filtering, and detection.
The real-time on-board processing of the data is aimed at continuously monitoring the FOV for collision threats. If a threat is
detected then the SAA system immediately informs the remote
pilot of the threat by sending a warning message though the ScanEagle's data link to the GCS. When the warning message has
been received the SAA terminal sounds an audible warning, which
alerts the remote pilot that a collision threat has been detected. The
remote pilot may then manoeuvre the ScanEagle to avoid the collision threat.
The main aim of the FT&E campaign was to test the detection performance of the SAA system. The comprehensive testing
of avoidance manoeuvres was left for future investigation.

Hardware
A low-power consumption single board computer (SBC) is at the
heart of the SAA system. The SBC has a powerful multicore parallel graphics processing unit (GPU), which is capable of performing complex image processing in real-time. The SBC managed the
data logging to the solid state drive (SSD) and data link communication. The SBC also controlled the triggering of the GPS/INS
system and the EO camera using a general-purpose input/output
(GPIO) interface.
A highly-integrated low-power consumption GPS/INS system
was selected to provide accurate measurements of the pitch, roll,
and heading of the aircraft. These measurements were used for real-time image stabilisation. Effective image stabilisation has been
found to be critical to the performance of the detection algorithm
[9].
A five megapixel commercial-off-the-shelf (COTS) camera
was selected for the SAA system. The camera used an IEEE 1394
(FireWire) serial bus interface to connect to the SBC. The combination of the camera and the lens sets the FOV of the SAA system
and the area imaged by each pixel. The area imaged by each pixel
is one parameter that sets the detection range of the SAA system
[11]. To explore this range dependence, two lenses were tested during the FT&E campaign: a 5 mm lens, which had a processed FOV
of 40° × 28°, and an 8 mm lens, which had a processed FOV of
26° × 18°.
A custom ScanEagle interface board was created for the SAA
system. The interface board provided a means of integrating a new
payload, such as the SAA system, with the existing ScanEagle system. Specifically, the interface board enabled the control of power
8

to the SAA system and enabled communication with the SAA terminal from within the ScanEagle's GCS.
Three-dimensional (3-D) models of all of the hardware systems and components were used to design rapid prototypes of a
custom ScanEagle mount for the SAA system. These 3-D printed
prototypes were used to check and adjust the initial design of the
mount. The final design enabled the manufacture of a robust mount
that allowed the SAA system to operate through all phases of flight
onboard the ScanEagle.

Software Architecture
The original SAA software baseline was developed and tested onboard a Cessna 172 [9], [10], [12]. The core software modules for
the SAA system are responsible for configuring the sensors, image
stabilisation, collision-threat detection, and data logging.
The transition of the SAA system to the ScanEagle platform
introduced some important changes to the software baseline. The
SWaP constraints limited the computation capabilities of the SAA
system. This resulted in an image processing frame rate of 9 Hz,
instead of the 15 Hz used in earlier research [10]. This decrease in
frame rate appeared to have no effect on the SAA system's detection performance.
Two user interfaces were added to the original software architecture. The SAA system control terminal was created to enable the
control and monitoring of the SAA system from the ground. The
SAA terminal was developed and installed within the ScanEagle's
GCS. The SAA terminal controlled power to the SAA system. The
SAA terminal also received aircraft detection warnings and other
status information from the SAA system.

Image Stabilisation
Image stabilisation refers to the real-time rotation and translation
of the images captured by the EO camera to compensate for image jitter caused by aircraft motion. Sources of aircraft motion that
require image stabilisation include aircraft vibration, wind gusts,
and air turbulence.
The role of the GPS/INS system was to measure the real-time
attitude of the aircraft. This information was then used to stabilise
the images from the EO camera. The performance of the GPS/INS
system sets the accuracy of the attitude measurements and, as a
consequence, the residual jitter level. Residual jitter is the motion
of the background features between consecutive stabilised image
frames. Thus, the GPS/INS system was carefully selected to minimise the residual jitter.
Previous research has shown that the residual image jitter has
an important effect on the detection performance of the SAA system [9]. Three levels of jitter were considered: low, moderate, and
extreme. It was shown that above a moderate level of jitter (three
pixels), the SAA system is not expected to detect any collision
threats.
The area imaged by a pixel is set by the combination of the
camera sensor and the lens. If the focal length of the lens is increased then a corresponding increase in the accuracy of the attitude measurements is required to keep the residual jitter constant.
Thus, a GPS/INS system was selected that could provide the re-

IEEE A&E SYSTEMS MAGAZINE

SEPTEMBER 2016

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine September 2016