The Magazine of IEEE-Eta Kappa Nu October 2017 - 24

FEATURE
the simulated mission by returning home to the launch
point and lands for recovery.

III. Onboard Electronics and Flight-Control
System
The onboard electronics and flight-control system
is responsible for the autonomous functionality and
image-capture capabilities of the UAS. These two main
electronic systems are illustrated in Figure 2.
The flight-control subsystem encompasses the flight
electronics and autonomous functions. The core of
this subsystem is the Pixhawk flight controller which
integrates several external sensors for navigation and
stable autonomous flight. The motors and control
surfaces are also part of this subsystem and consumes
most of the power. Power is supplied by two 14.8-V
batteries with 16-Ah capacities. This yields an overall
flight time of 30 minutes to an hour.
The image-capture subsystem's primary purpose is to
handle the capturing, processing, and transferring of
images to the GCS. The subsystem is comprised of a
camera, an onboard computer, and a WiFi radio. The
onboard computer is connected to the camera which
sends the camera capture commands. The images are
then offloaded to the onboard computer in real time.
The radio is connected to the GCS over WiFi and forms
a wireless bridge between the GCS and aircraft. The GCS
then downloads the images from the onboard computer
and stores the images for further use.

IV. Imaging System
The image-processing system is responsible for analyzing
target characteristics and estimating the geolocation
of targets. During the simulated SAR mission, targets
have five characteristics: background color, shape, letter
color, letter, and letter orientation. It must be sufficiently
lightweight to be included onboard the aircraft yet
powerful enough to perform [11].
Images are first captured by a ground-facing Nikon
Coolpix A camera controlled by a Raspberry Pi through
the gPhoto2 interface library. A script on the Raspberry Pi
triggers the camera to take pictures every four seconds
to ensure proper search-area coverage. Each image is
processed on the Raspberry Pi for targets and sent to the
GCS for further analysis.
A custom C++ program using the OpenCV library is used
to analyze captured images [12]. The program detects
24

Fig. 2. Onboard electronics and flight-control system diagram

Fig. 3. Image processing workflow

targets in an image, crops the target, and identfies the
shape, background color, and letter color. It saves this
cropped target to a folder of targets. After the program
is finished, the GCS crew manually checks the folder to
confirm the target and its characteristics. Figure 3 depicts
the workflow used by the image-processing system.
For the geotagging task, the GPS coordinates of the image
center are estimated using the aircraft's GPS coordinates,
altitude, pitch, roll, and yaw. These values are fetched from
the telemetry log file. An excel spreadsheet calculates the
estimated GPS coordinates of the image center.
The C++ program was designed to filter out any image
that has no targets detected, thus reducing the total
number of raw images. This method is completed by
using the OpenCV SimpleBlobDetector class [13], a
basic algorithm that creates clusters of pixels based
on color difference and thresholds. Images identified
to contain possible targets are cropped to focus on
the target. The next step involves applying Canny edge
detection [14] to each cropped target. The algorithm
takes the edges of each target and approximates the
closest polygon matches. These approximated polygons
are compared with a feature database of possible target
shapes. If the target matches a valid shape, it is put
through OpenCV kMeans clustering [15], a mathematical
www.hkn.org //

THE BRIDGE
http://www.hkn.org
Table of Contents for the Digital Edition of The Magazine of IEEE-Eta Kappa Nu October 2017
