IEEE Geoscience and Remote Sensing Magazine - March 2020 - 16

	

v CR =

3
2N

1 - c2
rc , (8)

where v CR is the standard deviation of the pixel-offset error, N is the number of samples in the estimation window,
and c is the coherence. For the typical incoherent cross correlation, the error is larger by at least 2 and, for lower coherence, is even worse [50].
Therefore, subpixel accuracy is limited, and, in general,
we can assume a 0.1-1 pixel accuracy. However, a higher
accuracy can be achieved when, similar to PSI, point scatterers are used. In this case, the accuracy can be estimated
with (see [50])
	

3
v point = r

1
, (9)
SCR

APPLICATION DEVELOPMENT
Methods are determined first, and applications follow. Our
three phases follow this trend, and the direction of methodological developments may give us hints about future
applications.
In the first phase, the lack of data availability hindered
most commercial applications. Scientific applications were
created but were mostly reduced to case studies. Nevertheless, the methods devised in this phase and tested within
these case studies built the foundation for the following
fast growth of applications outside of the scientific domain.
Early scientific applications are found, for example, in volcanology [12], [63] and urban areas [64] but in the form of
case studies, due to limited data availability. The first commercial company dedicated to satellite SAR ground-motion
estimates, TRE, chaired by Dr. Alessandro Ferretti, opened
during in this period, in 2000.

36°03′36″ N

36°03′54″ N

where SCR is the SCR of the target. For targets with an
SCR of 20 dB, the accuracy is about 1/20 of a pixel. Point
targets with high SCR can, therefore, be positioned with
a high subpixel accuracy, which is used, for example, in
image coregistration [51] and also for surface motion with
the point-target-offset tracking approach [52]. However,
as this method requires point targets with high SCR, it is
not suitable for all areas. Point targets with high SCR are
generally also suitable for D-InSAR-based methods; thus,

point-target-offset tracking is often applied in applications
in which the velocity is too high for D-InSAR-based
methods or for the estimation of motion in the azimuth
direction (Figure 9). Fast-moving landslides are a typical
application (Figure 10).
The subpixel positioning of point targets can further
be used for the precise positioning of targets [53], [54].
Using a stereoconfiguration, this SAR geodesy approach
allows the precise absolute positioning to decimeter and
subdecimeter precision [55]-[57]. This can be used, for
example, for the absolute positioning of PS or SAR tomography point clouds [58], [59]. These precisely measured
PSs can also be used as a replacement for ground-control
points (GCPs) [60], [61], which can, via SAR optical-image
matching, be used for GCP-free georeferencing of remote
sensing images [62].

300 m
cm/y
0
-20
-40
-60
-80

FIGURE 9. A linear-velocity map of the Beijing Capital Airport,

processed with SqueeSAR. (Source: TRE-Altamira; used with
permission.)
16

36°03′18″ N

LOS

101°10′30″ E

Flight

in geology [46], landslide detection and monitoring [47],
[48], and estimation of surface deformation caused by
nuclear tests [49].
The general pixel-offset tracking method is based on
normalized cross correlation; however, it can also be based
on complex coherence. For coherent cross correlation, the
achievable accuracy is higher with (see [50])

101°10′48″ E 101°11′06″ E

FIGURE 10. The landslide motion estimated from point-targetoffset tracking at the Lasiwa landslide. (Taken from [68].)
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2020
IEEE Geoscience and Remote Sensing Magazine - March 2020

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2020
