Aerospace and Electronic Systems Magazine July 2017 Tutorial XI - 2

Tutorial:

DOI. No. 10.1109/MAES.2017.160155

Introductory View of Anomalous Change Detection in
Hyperspectral Images Within a Theoretical Gaussian
Framework
Nicola Acito, Marco Diani, Accademia Navale, Livorno, Italy
Giovanni Corsini, Università di Pisa, Pisa, Italy
Salvatore Resta, Italian Navy

Exploitation of temporal series of hyperspectral images is a
relatively new discipline that has gained a lot of attention from
the image processing scientific community. In this paper, we
consider the specific problem of anomalous change detection
(ACD) in hyperspectral images and discuss how images taken at two different times can be processed to detect changes
caused by insertion, deletion, or displacement of small objects
in the monitored scene. We introduce the ACD problem using
an approach based on the statistical decision theory and we derive a common framework including different ACD approaches. Far from being inclusive of all the methods proposed in the
literature, this tutorial overview places emphasis on techniques
based on the multivariate Gaussian model that allows a formal
presentation of the ACD problem and the rigorous derivation
of the possible solutions in a way that is both mathematically
more tractable and easier to interpret. The unification of different approaches under a single rigorous statistical scheme
provides both a tutorial overview of ACD techniques, and a
useful instrument for researchers already familiar with the
ACD problem. Dedicated preprocessing methods aimed at improving the robustness of the ACD process are also discussed.
Real data are exploited to test and compare the presented
methods, highlighting advantages and drawbacks of each approach. The tutorial aspect of the paper has suggested the use
of a freely available data set. This should hopefully motivate
the interested reader to experiment with the processing methods and performance evaluation chain presented herein.

Authors' addresses: N. Acito, M. Diani, Dipartimento Armi
Navali, Accademia Navale, viale italia 72, Livorno, 57121 Italy.
E-mail: (n.acito@iet.unipi.it);
G. Corsini, Dipartimento di Ingegneria dell'Informazione,
Università di Pisa, Pisa, Italy;
S. Resta, Italian Navy and Dipartimento di Ingegneria
dell'Informazione, Università di Pisa, Pisa, Italy.
Manuscript received July 11, 2016, revised January 19, 2017,
and ready for publication February 20, 2017.
Review handled by W. D. Blair.
0885/8985/17/$26.00 © 2017 IEEE
2

INTRODUCTION
Change detection is a challenging task aimed at detecting pixels
that are substantially different in multitemporal images of the
same scene. Change detection has been effectively employed in
several remote sensing applications including both environmental
monitoring (such as geology, land cover and hydrology [12], [23])
and defense and surveillance applications (e.g. tracking of vehicles, identification of potential threats, damage assessment [20],
[51]). In this framework, change detection based on the analysis
of multitemporal hyperspectral imagery (HSI) is a recent emerging technology that is gaining increasing interest in the research
community because of the potential benefits expected in practical
applications. Change detection in HSI is accomplished by checking for changes in the spectrum of spatially coincident pixels/areas
in the different images. Such spectral changes may be caused by
object motion or by a change in the object status as, for example,
in the case of camouflaged objects.
Change detection has many aspects closely related to the field
of image processing. Works presented in the literature witness that
change detection has indeed benefited from studies on image registration [26], [30], [62], background modeling [10], [33], optical
flow [9], [60], and nonuniformity artifacts correction [59]. Furthermore, different solutions to the change detection problem have
been proposed in the remote sensing literature both for environmental monitoring and surveillance applications [11], [25], [38]
from spaceborne and airborne platforms [15], [21], [39].
Broadly speaking, a change detection algorithm takes the image sequence as input and generates a single image, called the
change statistic, whose highest values are associated to the relevant
changes. The challenge is designing a change detection strategy
mostly insensitive to pervasive changes, such as those induced by
sensor noise, contrast, brightness or focus differences, shadow and
camera motion, or atmospheric or even seasonal changes.
The distinction between pervasive and relevant changes, which
is fundamental for the choice of the proper change detection strategy, depends on the application of interest. Thus, in this work, we
focus on the detection of small changes resulting from insertion,
deletion, or movement of small size objects (generally man-made),

IEEE A&E SYSTEMS MAGAZINE

JULY 2017, Part II of II

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine July 2017 Tutorial XI