IEEE - Aerospace and Electronic Systems - July 2021 - 26

Feature Article:
DOI. No. 10.1109/MAES.2021.3065069
A Survey of Artificial Intelligence Approaches for
Target Surveillance With Radar Sensors
Andrea Wrabel, Roland Graef , and Tobias Brosch, Hensoldt Sensors, Ulm
89077, Germany
INTRODUCTION
Today's fast-paced technological progress presents companies
with the challenge of identifying and harnessing the
potential in new approaches before their competitors. In
such a competitive environment, companies cannot avoid
finding new ideas and solutions to meet the constantly
increasing technological requirements. In this context, the
use of artificial intelligence (AI) plays an increasingly
important role in the search for innovative approaches and
technologies [1]-[3]. Especially in recent years, this area
has experienced rapid development and is attracting great
interest from both industry and science. Hence, the number
of new AI-related patents increased from approximately
23 000 in 2008 to approximately 78 000 in 2018 according
to IPlytics [4]. In addition, there was a substantial increase
in AI-related paper publications according to the China
Institute for Science and Technology Policy at Tsinghua
University [5]. While only about 24 000 AI-related papers
were published in 1997, the number of published papers in
2017 could already reach a value of approximately
135 000. In the context ofAI-related investments, the funding
of AI startups experienced a huge growth from USD
$4.25 billion in 2014 to approximately USD $26.6 billion
in 2019 according to a survey from CB Insights [6].
The potential of AI algorithms is particularly evident
in finding patterns in large amounts of data [7], [8]. This
not only offers the possibility to draw conclusions based
on data analysis, but also to advance the automation of
various processes and tasks [1], [2], [3], [9]. Hence, AI
reveals a high potential in a broad range of possible applications
in various fields [3], [8]. Therefore, it is not surprising
that AI approaches might be helpful to support
Authors' current address: Andrea Wrabel, Roland
Graef, Tobias Brosch, Hensoldt Sensors, Ulm 89077,
Germany (e-mail: roland.graef@hensoldt.net).
Manuscript received March 19, 2020, revised August 12,
2020; accepted November 14, 2020, and ready for
publication March 8, 2021.
Review handled by Chee-Yee Chong.
0885-8985/21/$26.00 ß 2021 IEEE
26
various tasks in radar data processing and could possibly
further refine existing approaches or even serve as a powerful
alternative to conventional techniques. In this regard,
Table 1 provides an overview of common AI approaches
as well as corresponding descriptions and exemplary
applications in radar data processing. For further exemplary
application not listed in Table 1 we refer to
Majumder et al. [10]. Since target surveillance is a major
application field of radars in civilian as well as in military
contexts, we focus on this type of radar sensor approaches
within this article. With this in mind, we aim to review the
most important aspects in past and recent research developing
AI approaches for target surveillance using radar
sensors. Indeed, the recent interest in AI is not new so that
we aim at revealing the potential of past approaches,
which could be with today's computational capabilities,
suitable for application in practice. Additionally, current
hot topics arising from recent research shall be presented.
Hence, the goal of this article is to give an overview of
the following three major research streams exploiting the
capabilities of AI approaches for target surveillance with
radar sensors: 1) clutter identification, 2) target classification,
and 3) target tracking. Indeed, these are pivotal assets
for adequate target surveillance. Since clutter (i.e., undesired
signal reflections) can strongly impede the detection
of targets, its identification and subsequent suppression is
of great importance. Further, a correct target classification
can contribute, particularly in military contexts, to a successful
prevention of potential hazards. Finally, target
tracking, representing the last link in the conventional
chain of radar data processing, requires special attention
as it offers the key point for sensor data fusion. In this
regard, we do not present a full survey, but aim at providing
the reader with some of the most relevant information
to follow up on any of these major research streams covered.
Hence, we present a representative sample of literature
over the last decades regarding approaches for target
surveillance based on radar sensors. Thereby, we discuss
the potential of the approaches for application in practice
and provide insight into our own findings.
The remainder of this article is organized as follows:
In the next section, we review the most representative
IEEE A&E SYSTEMS MAGAZINE
JULY 2021
https://www.orcid.org/0000-0002-0917-2541 https://www.orcid.org/0000-0002-0917-2541 https://www.orcid.org/0000-0002-0917-2541 https://www.orcid.org/0000-0002-0917-2541 https://www.orcid.org/0000-0002-0917-2541
IEEE - Aerospace and Electronic Systems - July 2021

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - July 2021
