Aerospace and Electronic Systems - March 2019 - 18

Feature Article:

DOI. No. 10.1109/MAES.2019.2900178

Robust Cooperative Target Detection for a VisionBased UAVs Autonomous Aerial Refueling Platform
via the Contrast Sensitivity Mechanism of Eagle's Eye
Haibin Duan, Long Xin, Shanjun Chen, Beihang University, Beijing, China;
Peng Cheng Laboratory, Shenzhen, China

INTRODUCTION
Unmanned aerial vehicles (uavs) have been widely utilized to perform various tasks in civilian and military,
such as target detection and tracking [1] surveillance
[2], [3], forest firefighting [4], formation flight [5], [6],
vision-based navigation [7], and collision avoidance
[8], [9]. one important challenge that limits the uavs is
the lack of capability of aerial refueling. autonomous
aerial refueling (aar) extends the range, payload, and
endurance time of aircrafts, being the amplifier of air
power for uavs [10].
Currently, aerial refueling can be classified into the
probe-and-drogue and the refueling boom. The former is
the standard for the United States Navy. The refueling
boom is used by the United States Air Force developed by
Boeing, which is also the research object of this paper.
The premise of AAR is to obtain the relative pose information of the receiver UAV and the tanker aircraft. Compared to the classical sensing technologies, such as inertial
navigation system (INS), differential GPS or intra flight
data link (IFDL), the vision-based measuring system is
increasingly popular in recent years [11]-[15]. In order to
solve the problem of pose estimation, the VisNav sensor
is used by Valasek et al. [16], and a vision-based strategy
is also proposed by Martınez et al. [12] for the probe and
drogue system. For the boom refueling system, through
the extended Kalman filter (EKF), Mammarella et al. [17]

Authors' current address: Haibin Duan, Long Xin,
Shanjun Chen, State Key Laboratory of Virtual Reality
Technology and Systems, Beihang University, Beijing
100083, China; Peng Cheng Laboratory, Shenzhen
518000, China. E-mail: (hbduan@buaa.edu.cn).
Manuscript received February 5, 2018; revised July 19,
2018. ready for publication February 10, 2019.
Review handled by M. Cardinale.
0885/8985/19/$26.00 ß 2019 IEEE
18

combined the position information from GPS and a
machine vision (MV) based system to provide a reliable
relative position estimation of the receiver UAV, and the
optical sensor of the visual snake is utilized by Deobbler
et al. [18] to solve the problem. Besides, Duan et al. [19],
[20] established a hardware-in-loop simulation platform
on the basis of computer vision; Gao et al. [21] utilized
low rank and sparse decomposition with multiple features
to solve the problem of drogue detection. A visual detection and tracking algorithm using the shape feature of the
drogue is designed by Yin et al. [22], and a real-time position sensing system based on MV is also designed by
Dell'Aquilla et al. [23] to solve the AAR problem.
Although the vision-based measuring system for the pose
information has already been studied in a lot of literature,
its implementation for AAR between UAVs remains a
challenge in a real environment.
In the marker detection, some markers may be sheltered while disturbance points may be detected as feature
points. In order to improve the accuracy of marker detection, the contrast sensitivity mechanism, inspired by the
properties of the contrast sensitivity function (CSF) of
eagle's eye, is utilized to extract the cooperative target.
The specific biological explanations of visual processing
have been given by Harmening [24]. And the distribution
of neurons projecting from the retina and visual cortex to
the thalamus and tectum opticum of the barn owl was also
researched by Bravo et al. [25]
The retina of the eagle is constituted by the typical
cell-layers found in other vertebrates. They have evolved
the specific sensory and morphological adaptations to the
small targets [26]. An eagle flying at an altitude of 1000 ft
is able to determine the prey [27]. Additionally, there are
two foveae in eagle's eye: the deep fovea the shallow
fovea. Compared to the shallow fovea, the deep fovea
with a pit-like cross section is deeper, and the receptor
density is higher than that of the shallow fovea [28].
Figure 1 shows the physiological structure of the eagle's

IEEE A&E SYSTEMS MAGAZINE

MARCH 2019
Aerospace and Electronic Systems - March 2019

Table of Contents for the Digital Edition of Aerospace and Electronic Systems - March 2019
