Spatiotemporal Change Detection Based on
Persistent Scatterer Interferometry:
A Case Study of Monitoring Building Changes
C. H. Yang, B. K. Kenduiywo, and U. Soergel
Abstract
Persistent scatterer interferometry (PSI) detects and analyses
PS points from multitemporal SAR images for scene deforma-
tion monitoring. We propose a novel technique to identify
disappearing and emerging PS points, which are regarded
as building changes in cities. A spatiotemporal analysis is
implemented as both spatial position and occurrence time are
obtained for each change point. We first estimate each pixel’s
temporal coherences in different image subsets. Computed
from temporal coherences, a change index sequence is intro-
duced to quantify each pixel’s probabilities of being change
points at different times. All pixels’ change indices are then
utilized to extract change points by a global, automatic, and
statistical-based scheme. We then eliminate blunders by a
spatial filtering. Finally, the occurrence times of the change
points are detected based on the temporal variation in their
change index sequences. We implement a simulated data test
to validate and assess our method. Using TerraSAR-X images,
our real data test successfully recognizes steady, disappear-
ing, and emerging buildings in Berlin, Germany within 2013.
Introduction
The continuous rise in population and economic growth has
led to urbanization across the world coming along with frequent
building changes, e.g., construction, in a built-up environment.
Monitoring such changes is important for city management,
urban planning, updating of cadastral maps, etc. (Gamba, 2013;
Marin
et al
., 2015). Remote sensing offers a fast and cost-effec-
tive mapping of large areas compared with conventional field
surveys. Particularly, spaceborne synthetic aperture radar (
SAR
)
sensors provide radar images captured rapidly over vast areas
at fine spatiotemporal resolution. The
SAR
sensors are weather
independent and have a day-and-night vision ability, which
guarantees images with a high temporal density. These capabili-
ties make
SAR
suitable for monitoring events.
Many time series analysis methods using multitemporal
SAR
images have been proposed for urban monitoring. For instance,
persistent scatterer interferometry (
PSI
) (Costantini
et al
., 2008;
Crosetto
et al
., 2005 and 2016; Ferretti
et al
., 2000, 2001, and
2011; Hooper
et al
., 2004; Kampes, 2006) detects persistent
scatterer (
PS
) points, which are characterized by strong, stable,
and coherent radar signals throughout a
SAR
image sequence.
In principle,
PSI
eventually derives for each
PS
a set of attri-
butes, such as temporal coherence, line-of-sight (LoS) velocity
(mm/year level), topography height, and geographic position.
These attributes are then used for scene monitoring. A signal
sequence of a
PS
point is modeled to maintain coherence dur-
ing the entire acquisition period of a
SAR
image stack. Accord-
ingly, a scene of interest covered with
PS
points is assumed to
be steady and free of big changes. For example,
PSI
works well
in monitoring of built-up cities because the regular and station-
ary substructures of buildings cause high
PS
density. However,
if the substructures or even entire buildings disappear due to
construction, the corresponding semi-
PS
points are discarded
in the initial screening of
PSI
processing. In other words, big
changes cannot be revealed by common
PSI
.
Some previous works (Ansari
et al
., 2014; Brcic and Adam,
2013; Ferretti
et al
., 2003; Novali
et al
., 2004) aim at detecting
semi-
PS
points, which disappear or emerge at arbitrary times,
by looking for abrupt amplitude changes of pixels along
SAR
image stack. Indeed, the amplitude-based thresholding
methods (Adam
et al
., 2003; Crosetto
et al
., 2003; Ferretti
et
al
., 2001; Lyons and Sandwell, 2003; Werner
et al
., 2003) are
commonly used to choose
PS
candidates, e.g., by means of
amplitude dispersion (Ferretti
et al
., 2001). Here, high and
stable amplitudes indicate potential
PS
points. However, we
might miss those low-amplitude
PS
points even their signals
are permanently coherent and stable. Consequently, low-am-
plitude semi-
PS
points cannot be recognized either. To avoid
such loss, an alternative strategy (Hooper
et al
., 2004) exploits
temporal coherence, an indicator of temporal phase stability,
for
PS
identification. This strategy is adapted and extended in
our method for change detection.
In this study, we propose spatiotemporal change detection
based on
PSI
to detect disappearing and emerging semi-
PS
points along with their occurrence times. The term “spa-
tiotemporal” refers to changes that occur over geographical
space and are detected over time. We distinguish and label
these two scenarios as disappearing big change (
DBC
) and
emerging big change (
EBC
) points. To begin with, multitempo-
ral
SAR
images are divided into several subsets by a sequence
of break dates (
bd
). The temporal coherence of each pixel
in an image set is estimated using a standard
PSI
processing.
Temporal coherence is modeled to be proportional to phase
stability and thus serves as an indicator of a
PS
point. The key
idea of our approach is to derive a change index sequence for
each pixel from its temporal coherence estimates spanning
different periods. An automatic thresholding is applied to the
change indices to determine the final change points. We then
eliminate blunders by a spatial filtering, which is a crucial
element for a spatiotemporal analysis. Finally, we check the
evolution of a change index sequence to identify the probable
C. H. Yang and U. Soergel are with the Institute for
Photogrammetry, University of Stuttgart, Geschwister-Scholl-
Str. 24D, 70174 Stuttgart, Germany (
).
B. K. Kenduiywo is with the Department of Geomatic
Engineering and Geospatial Information Systems, Jomo
Kenyatta University of Agriculture and Technology, P.O. Box
62000-00200, Nairobi, Kenya.
Photogrammetric Engineering & Remote Sensing
Vol. 84, No. 5, May 2018, pp. 309–328.
0099-1112/18/309–328
© 2018 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.84.5.309
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2018
309
