The inter-orbital distance causes some phase difference on images. This information is combined with precise
orbital information to determine the elevation and shift of each pixel. In order to obtaining of absolute DEM at
least one three-dimensional ground control point is required.
The processes of interferometric terrain extraction and surface shift estimation are quite complicated however
InSAR/DInSAR processor uses sophisticated techniques, such self-automatic images co-registration, adjustment of
orbital parameters, interferogram generation, phase filtering and phase unwrapping which are designed in manner
to simplify their usage.
The output DEM or surface shift map are referenced to WGS84 ellipsoid and allocated in geographic projection.
Accurate coregistration of two radar images to 1/8 pixel is necessary for precise
interferogram calculation. The main coregistration aim is to transform slave
image in master image geometry with help of affine transformation. It is assumed
that satellite orbits is parallel and affine transformation is enough for slave
image geometry adjustment.
Interferogram and coherence calculation
Interferogram is a result of complex multiplication of master image and complex
conjugate to slave image.
In result phase differences image of a surface is formed. Besides of this, simultaneously
flat Earth phase correction operation is executed. Flat phase correction includes
operations of phase correction in range and azimuth directions. Flat phase correction
in range is necessary because the phase difference image contains the information
both about height of object and about distance up to object. Flat phase correction
in azimuth is necessary because the satellite orbits is not parallel during
master and slave images obtaining.
The choice of a mode "Precise"
means calculation of coefficients of compensation for each image pixel individually.
Thus time for realization of operation is increased.
In a mode "Fast"
the coefficients of compensation calculated for the central column and the central
line of a picture are used.
In a mode "No flattening" flat Earth phase correction operation will
not be executed.
Simultaneously with interferogram calculation, coherence calculation is carried
out. Coherence is a parameter describing a synchronization of values of the
main and slave images. More close to 1 value of coherence in pixel speaks about
a good signal/noise ratio for interferogram image in that pixel.
0.0 - 0.3 - poor. Typically for interferogram areas where there are phase discontinuity.
0.3 - 0.5 - low. Typically for interferogram areas with changed between observations
0.5 - 0.7 - good.
0.7 - 1.0 - excellent.
Filtration of interferogram is performed in order increase the signal-noise
ratio. Speckle noise, which is inherent for the coherent wave systems, presents
on both the radar images and the interferogram. Speckle noise is not only decreases
the visual quality of interferogram but additionally considerably complicates
and worsens quality of work of phase unwrapping algorithm, induces the errors
in absolute phases matrix.
It’s necessary to note that filtration decreases the spatial resolution on interferogram.
Therefore the filter parameters should be selected taking into account this
fact. It’s desirable to avoid too intensive interferogram filtration.
The following filters are accessible to use:
Average filter. Most simple
of accessible filters. From the mathematical point of view this filter is equivalent
to construction of a plane on some set of points by the least squares method.
Works quickly but filtration quality not always high. The filtration the more
strongly, than is more size of a filtration window. The filtration with a window
of the size 1x1 means absence of a filtration.
Spectral adaptive filter. At
work uses two dimensional Fourie transformation. Varying Min and Max values
of parameter of filtration intensity it is possible to achieve necessary quality
of a filtration. The more value of parameter of intensity, the more strongly
will be filtration.
Phase unwrapping procedure have to solve a problem of phase ambiguity. Before
phase unwrapping interferogram contains phase values in limits 0-360 degrees
but the knowledge of absolute phase values is necessary. The problem is reduced
to addition of the necessary number of phase cycles to each phase measurement.
The decision of a problem is complicated by presence on interferogram phase
noise and so called residues, that makes the decision of a problem difficult
theoretically and expensive computational. There is a set of methods of the
phase unwrapping problem decision.
The following methods are accessible to use:
Unweighted phase unwrapping.
The most simple algorithm on the base of the least squares method.
Picard iteration method. Algorithm
on the base of the least squares method with use of Picard iterations method
for decision of system of linear algebraic equations.
Conjugate gradients method.
Algorithm on the base of the least squares method with use of conjugate gradients
method for iterative decision of system of linear algebraic equations.
Growing pixels method. This
algorithm realizes the local approach in phase unwrapping. Phase unwrapping
will be carried out from the unwrapped pixel the nearest not unwrapped. Quality
of algorithm work is checked by threshold tests.
There is also an opportunity of the passing of a phase unwrapping stage. It
may be useful in case of area with insignificant differences of heights, when
on interferogram change of a phase does not cross border of 0-360 degrees.
Absolute phase values to height recalculation
Values of an absolute phase can be recalculated both in relative or in absolute
values of height.
For recalculation of an absolute phase in relative values of height it is only
necessary to specify a point concerning which recalculation will be made. Result
of recalculation is the image for which the difference of heights between two
neighbour pixels is kept.
For recalculation of absolute phase values in height
values it is necessary to specify several ground control points, heights in
which are authentically known. It is desirable, that ground control points were
evenly distributed on the image. Result of recalculation is the image of the
heights above reference ellipsoid.
At this stage geocoding of matrix of the heights received at a stage of recalculation
of absolute phase values in height is carried out. The relief thus is translated
in geographical system of coordinates: latitude-longitude-height above reference
ellipsoid. The opportunity of a choice of the geometrical sizes of an output
file grid on latitude and longitude is given. In result the file representing
digital elevation model turns out.
Digital elevation model
Differential SAR interferometric data processing
Interferometric Processor software allows to process SAR data, acquired in interferometric mode (a 3-path and a 4-path interferometry). The differential SAR interferometric data set consists of 3 or 4 images (regarding PHOTOMOD Radar), each of them makes interferometric pair to any other image data of this set. Processing images of such set in pairs makes possible to get the differential interferogram, the picture of a phase changes (interferograms) for a some period of time. Considering a viewing geometry of all data of the set, you can re-calculate phase values of differential interferogram into field of relief displacement, occurred between observations of the processed pairs. Such changes can be caused both by vertical and horizontal motions of an underlying surface. As values of the differential interferogram directly depend on SAR wavelength, displacements on the surface can be defined in millimeter scale.
The processing of differential SAR interferometric data in the processor module is made consistently using the same tabs, as a processing of interferometric pairs with following features.