IEEE Geoscience and Remote Sensing Magazine - March 2020 - 25

	

z flat =

4r <
$ B , (2)
m

where B < is the parallel baseline. z dem in (1) is the topographic phase contribution, proportional to the perpendicular baseline [25]
4r
	
z dem =
$ h $ B =, (3)
mR sin i
where R is the range between the sensor and the target, i
is the local incident angle, h is the terrain height, and B =
is the perpendicular baseline. z def in (1) is the deformation phase component (note that it is along the line-of-sight
direction of the satellite [6]). z aps in (1) is the atmospheric
propagation delay component [also called the atmospheric
phase screen (APS)].And z noise in (1) is the random noise.
Specifically, for DInSAR processing, the observable interferometric phase z is the value wrapped between -r
and r:
	

shows the development of various time-series InSAR techniques. We classified current time-series InSAR techniques
according to different perspectives so that readers may
have a comprehensive understanding of various methods
and learn about the design ideas behind them. The classification of time-series InSAR techniques is based on the
key steps in their processing chain.
According to the type of measurement point utilized,
classification can be divided into methods based on PSs,
distributed scatterers (DSs), and the combination of
PSs and DSs. PSs are ground objects that can maintain
strong, stable electromagnetic scattering properties over
a long time sequence. PSs are mainly distributed in urban areas, such as houses, roads, and bridges [36]-[38].

U = z + k $ 2r, (4)

where U is the unwrapped phase and k is an integer value (called phase ambiguity). So the fringe pattern can be
seen in the interferogram [26]. The density of the fringes
depends on terrain change, baseline length, and signal
wavelength, and the clarity of the fringes reflects the
quality of the interferogram. Phase unwrapping, i.e., recovering the phase ambiguity of 2π, must be performed
[27]-[33]. Phase unwrapping has always been an important and challenging issue in the SAR interferometry
field [34].
The spatial baseline is an important parameter in SAR
interferometry processing. It is a vector pointing from S 1
to S 2, which can be decomposed into parallel baseline
component B < and perpendicular baseline component
B =, as shown in Figure 4. Accurate baseline estimation
has a vital impact on subsequent data processing. Baseline estimation with centimeter or even millimeter accuracy is necessary [35], and it is usually derived from
orbital vectors and refined using interferograms.
In DInSAR, the flat Earth and topographic phases are
subtracted from the interferometric phase z to obtain the
deformation signal [20]. The APS is usually ignored or removed by external data. In the case of time-series InSAR,
many interferograms can be generated, and the signal to
be processed changes from 2D (azimuth and range) to 3D
(azimuth, range, and time). Although many frameworks
are applied for time-series InSAR processing, they share
the same basic principle. In essence, the time-series InSAR technique is used to process the complex 3D signal to
obtain the information of interest, as shown in Figure 5.

S2

S1

P1

P2

FIGURE 3. An illustration of DInSAR observation geometry. S 1 and

S 2 are the master and slave acquisition, respectively. During the
observation interval, the target moves from P1 to P 2 . The cloud
indicates that the signal travels through the atmosphere.

S2

B
S1
B

⊥

B

TIME-SERIES INTERFEROMETRIC SYNTHETIC
APERTURE RADAR TECHNIQUES
CLASSIFICATION
Since the introduction of PSI, many organizations have
proposed their own time-series InSAR techniques. Figure 6
MARCH 2020

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

FIGURE 4. The parallel baseline B < and perpendicular baseline B = .

S 1 and S 2 are the master and slave acquisition, respectively.
25
IEEE Geoscience and Remote Sensing Magazine - March 2020

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