377
In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7,2010, IAPRS, Vol. XXXVIII, Part 7B
Interferometry, Small BAseline Subset, Coherent Pixel
Technique, which will be introduced chronologically in some
detail, together with the differences and relations between each.
2.1 Least Square database approach (LS approach)
After analyzing the phase stability of some man-made features,
S.Usai et al (Usai et al., 1997,1999,2000) presented a new
approach, known as Least Square approach (LS), for the long
term monitoring of terrain deformations with D-InSAR. This
method uses a database of interferograms, and by solving all the
deformation velocities as a unique least squares problem
provides a chronologically ordered sequence, describing the
evolution of the deformation pattern in time (Usai, 2002).
2.1.1 Main principle
The input for the least squares adjustment is a set y=[Ii,...,I N ] of
N unwrapped interferometric deformation maps(i.e., the
unwrapped interferograms are compensated for topographic and
flat-earth phases), all coregistered at subpixel level, generated
from M SAR images taken at days x=[d 1 ,..., d M ]. The day di
corresponding to the first image is taken as reference and the
deformations at each of the other (M-l) days relative to this day
considered as solutions of the problem:
y=Ax (1)
where in x the element d] not been considered.
In the system matrix A, each row corresponds to an
interferogram, while the columns correspond to the days. For
interferogram I k —Idi-Idj» the values on row k are all zero except at
columns i and j, being +1 and -1 respectively and A an
incidence like matrix, directly depending on the set of
interferograms generated form the M SAR images.
The unweighted least squares solution x of Eq.l is
straightforward:
x=(A T A)-'A T y Qx=(A T A)-‘ (2)
2.1.2 Some notes on LS approach
The LS approach has been applied to measure terrain
displacements in the period 1993-1999 at the Phlegrean Fields
(Naples, Italy), using a set of 20 ERS-1/2 SAR images, and 43
interferograms generated(Usai, 2002; Usai, 2003). The authors
made use of an external DEM to obtain the differential
interferograms and found that residual topography had caused
systemtic effects in the data. In addition, during the processing,
a closed-loop method was used to detect and remove image-
and interferogram-related biases. In fact, according to Usai,
(Usai, 2001) two kinds of biases can be identified: the image-
related ones, like for example those caused by atmospheric
disturbances; and the interferogram-related ones, i.e. those
which have been produced in the interferometric combination of
two images, most probably by phase-unwrapping errors.
2.2 Permanent Scatterer (PS)
The Permanent Scatterer technique, developed at the
‘Politecnico di Milano’ (Milan, Italy) (Ferretti et al., 1999;
2000; 2001), is the first of a family of similar advanced
interferometric techniques-Permanent Scatterer Interferometry
(PSI).
Given N+l images, a set of N differential interferograms is
generated with respect to a single master. High temporal and
normal baseline interferograms (affected by a high decorrelation
noise) are, thus, part of the dataset. The approach focused on
privileged image pixels that, even in these ‘extreme conditions’,
still exhibit a low noise term, thus the so-called Permanent
Scatterers. With PS technique, pixels are selected from the
study of its amplitude stability along the whole set of images
(typically >30). Therefore, the maximum resolution of the SLC
images is preserved (Ferretti et al., 2001)
After PS candidates selection, a linear model is adjusted to the
data to estimate the deformation linear velocity and possible
DEM errors for each PS Candidate. Then the atmospheric phase
screen (APS) for the master image and the nonlinear motion
contribution and APS for each image are computed through a
spatio-temporal filtering. After estimation and removal of all the
APS superimposed on the data, one can identify more PSs and
repeating the previous steps allows getting the whole
deformation time series and average LOS displacement rate of
every single PS, and a refining DEM with sub-metric precision
of the exact height of the object corresponding to the PSs.
2.3 Small BAseline Subset (SBAS)
The Small Baseline Subset (SBAS) approach, proposed by
Berardino et al. (2001; 2002), extends the Least Squares
approach (Usai, 2001; 2002; 2003; Lundgren et al, 2001) to the
case of multiple small baseline acquisition subsets. The key
point, in addition to the use of multi-look interferograms, is that
the data pairs involved in the generation of the interferograms
are carefully selected in order to minimize the spatial baseline,
thus mitigating the decorrelation phenomenon and topography
errors. The Singular Value Decomposition (SVD) method is
applied to link otherwise independent SAR datasets separated
by large baselines. The SBAS method was originally used to
investigate large scale deformations with spatial resolution of
about 100m* 100m, calculating the time sequence deformation
and estimating DEM error and the atmospheric artifact in a
similar way as PS. O.Mora et al (O.Mora et al., 2002) promoted
a complementary approach, utilizing two different sets of data
generated at low (muliti-look) and full resolution (single-look)
respectively, to monitor localized deformation. The former are
used to identify and estimate possible atmospheric phase
artifacts and low-wavenumber deformation patterns based on
SVD SBAS method or CPT(Mora et al., 2003); the latter to
detect, on the high-resolution residual phase components,
structures highly coherent over time like buildings, rocks, lava
structures, etc.
2.4 Coherent Pixels Technique (CPT)
Developed by O.Mora et al (2003), original CPT gained its first
use (Mora et al., 2001) in the long-term subsidence monitoring
of an area of small town in Spanish, choosing the temporal-
coherence as criterion for permanent scatterers selection only to
make flexible the SAR images requirement in PS. The results,
utilizing seven SAR images, turned out to be satisfactory and
coincided well with the DGPS measurements.
Recently, CPT has been improved (Blanco et al., 2007; Duque
et al., 2007) into an operational advanced technique for terrain
deformation mapping, in terms of linear and nonlinear
deformation extraction, robustness with DEM error, thus
allowing DEM refining, and atmospheric phase screen (APS)
removal. P.Blanco et al (2008) concluded this approach and
detailed the main steps, such as optimal interferogram sets
selection, coherent pixels selection, linear and nonlinear blocks
for a full deformation extraction. The related algorithms consist
of Delaunay triangulation and Minimum Spanning Tree (MST)
for best combination of interferograms selection, Conjugate
Gradient Method (CGM) for Phase Unwrapping, multi-layer for