IEEE Geoscience and Remote Sensing Magazine - March 2020 - 84

detect small or slow mining deformation (e.g., several to
dozens of centimeters per year) but are unable to identify large or fast mining deformation (e.g., several meters
annually) mainly because high deformation gradients in
mining areas likely cause discontinuous wrapped phases
in interferometric data, making it difficult to correctly
recover the real deformation pattern using the process of
phase unwrapping [136]-[138].
As stated previously, in recent years, pixel OT techniques
have been applied to estimate large mining deformation
(see the "1D LOS Mining Displacement Measurements"
section). However, the accuracy of the OT-derived displacement observations primarily depends on the spacing of the
SAR image pixels (~1/10-1/20 pixel size in general) [15],
[28]. As a result, large mining displacement measurements
are difficult to accurately generate using OT techniques for
those mining areas without available high-resolution SAR
images. At present, the pixel spacing of the available SAR
images over most mining areas is approximately 3-10 m,
which results in the accuracy of OT-derived displacement
observations being on the order of 0.3-1 m (taking 1/10 of
the pixel size as a reference). Such an accuracy is usually too
low for InSAR-based applications in mining engineering
(e.g., model parameter inversion and mechanism understanding). Hence, at present, it is a challenging task to accurately retrieve large mining displacement from SAR images.
FUTURE RESEARCH
Despite the fact that additional work is needed to address
these challenging issues in InSAR-based 1D/2D mining displacement retrieval, some promising attempts to improve
the previous methods of retrieving 3D mining displacements, inverse model parameters, and forward predict displacements are discussed in the following section. In addition, a few potential applications of InSAR measurements
not yet systematically focused on are outlined as well.
TOWARD ROBUST ESTIMATION OF 3D MINING
DISPLACEMENTS
As noted in the "SAR/InSAR-Based 3D Mining Displacement
Reconstruction" section, two types of methods (i.e., multitrack/multisensor and InSAR+model) have been developed
for estimating 3D mining displacements from InSAR measurements. Unfortunately, both types of methods have low
robustness, and studies to robustly estimate 3D mining displacements from InSAR measurements should be conducted
in the future. In fact, the low robustness of both the multitrack/multisensor and InSAR+model methods is due to the
lack of very redundant observations or constraints. More specifically, in terms of the multitrack/multisensor method, the
number of completely independent viewing geometries for
the current SAR sensors is two (i.e., ascending and descending orbits with right-looking angles) in most areas of Earth.
As a result, it is difficult to collect SAR data sets from three
synchronous independent viewing geometries, let alone redundant SAR images. In addition, the near-polar orbits of
84

current SAR satellites lead to insensitivity of the InSAR LOS
deformation to surface movement in the northern direction,
thus further affecting the robustness of 3D displacement estimation, especially in the northern direction [106].
Admittedly, several methods have been proposed to
improve the robustness of 3D displacement estimation
in recent years, e.g., by adding azimuth deformation observations generated by MAI or OT techniques [91], [139].
However, the practical applications of these methods are
limited because MAI and OT methods can accurately measure only azimuth displacement in a few mining areas with
high coherence (using MAI) or where high-resolution SAR
images are available (using OT) [22], [28]. As suggested by
Wright et al. in [87], it is theoretically possible to improve
the robustness of 3D displacement estimations for those
areas with SAR data sets from ascending and descending
passes using left- and right-looking radar. Unfortunately,
such a strategy is difficult to conduct in practice because
only right-looking view geometry is equipped in most current SAR sensors.
For the InSAR+model method, only three independent
equations are constructed for solving 3D mining displacements at a pixel (see the "Integration of InSAR Measurements With a Deformation Model" section for more information); this implies that redundant observations or
constraints are available to improve the robustness of the
3D displacement estimation. As a consequence, large errors
in deformation observations caused by the DInSAR (e.g.,
unwrapping) or OT (e.g., mismatching) processes would
propagate into the 3D displacement estimates [94]. Moreover, errors of the prior model used in the InSAR+model
method [see, e.g., (8)], caused by things such as ground fissures, would dramatically reduce the accuracy of 3D mining displacement estimates [99]. Although more deformation observations generated from TS SAR images can help
improve the robustness of 3D mining displacement estimation, the inherent errors caused by the prior model's uncertainties are difficult to mitigate [101].
Because the lack of effective redundant observations is
the main cause for the low robustness of existing multitrack/multisensor and InSAR+model methods for solving
3D mining displacements, a straightforward way to improve low robustness is to add more observations or constraints. Currently, a promising direction is to fuse multisource data sets (e.g., multipass SAR images, optical images,
and ground-based surveying measurements) with prior
models to solve 3D mining displacements.
TOWARD THE PROCESSING IMPROVEMENT OF
InSAR-BASED MODEL PARAMETER INVERSION
Although several methods have been developed for the inversion of mining model parameters and further prediction
using InSAR measurements, to date, some open problems
still exist in these methods, and processing improvement
of InSAR-based model parameter inversion is needed. For
instance, InSAR phase decorrelation generally causes the
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
