The filtering of layover areas in highresolution InSAR for the building extraction

1
The filtering of layover areas in high-resolution
InSAR for the building extraction
Conference on SAR Image Analysis, Modelling, and
Techniques V
Session 4 SAR Interferometry Applications
By David PETIT, Frédéric ADRAGNA and Jean-Denis
DUROU.
2
Components of phase
1. Theory of the spectral shift.
These components are supposed constant.
supposed negligible
varies with the local incident angle.
3
Construction phase
pixel size in range
Goodman model
construction phase
Fresnel representation
4
Phase difference
The phase difference between two small objects
?l
d
(f0 is the frequency of the wave)
With a slight change of sight
??l
d
5
The spectral shift
The construction phases are the same, only if
It is possible only if we suppose that in the
second case, the frequency is f0
General formula as introduced by C. Prati, F.
Gatelli et al
b? orthogonal baseline ? off-nadir angle d
distance ? local slope k 1 (monostatic case)
2 (bistatic case)
6
Sub-interferogram
Master image
Slave image
2. Separation of two slope interferograms.
?f
?f
FFT(range)
FFT(range)
f
f
B
FFT-1
FFT-1
Sub-interferogram
7
Several sub-interferograms
We can produce as many sub-interferograms as we
can select pairs of filtered images with the same
spectral shift.
?f
f
f
Sub-interferograms
8
Slope interferogram
Sub-interferograms
related to the slope
Signals
Sum Gradient subtraction
other contributions
Slope interferogram
9
Separating two slope interferograms
If ? is equal to 45 degrees, and if we want to
separate signals related to horizontal and
vertical planes
Master image
Slave image
?f
?f
slope selection
slope selection
f
f
Slope interferogram for a horizontal plane
Slope interferogram for a vertical plane
10
Unwrap the phase in a building layover
Interferograms
On a slide
On a model of building
3. Unwrapping in the building layovers.
20 m
shape of reference
reconstructed shape
11
Unfold the layover
Coherence for the slope interferogram in case of
vertical plane
Slope interferogram for horizontal plane
1
Threshold V
Slope interferogram for vertical plane
0
Stage
1
3
4
Shape of reference
2
12
Unfold the layover
Coherence for the slope interferogram in case of
vertical plane
Slope interferogram for horizontal plane
1
Threshold V
Slope interferogram for vertical plane
0
Stage
1
3
4
Shape of reference
2
13
Unfold the layover
Coherence for the slope interferogram in case of
vertical plane
Slope interferogram for horizontal plane
1
Threshold V
Slope interferogram for vertical plane
0
Stage
1
3
4
Shape of reference
2
14
Unfold the layover
Coherence for the slope interferogram in case of
vertical plane
Slope interferogram for horizontal plane
1
Threshold V
Slope interferogram for vertical plane
0
Stage
1
3
4
Shape of reference
2
15
Unfold the layover
Coherence for the slope interferogram in case of
vertical plane
Slope interferogram for horizontal plane
1
Threshold V
Slope interferogram for vertical plane
0
Stage
1
3
4
Shape of reference
2
16
Limitations
The spectral shift must meet certain conditions,
to allow using this technique of slope selection.
We can express these as a function of the
altitude of ambiguity Ea
Sd Sampling in range H Height of the building

• Resolution of the slope interfogram is all the
  more reduced that the selected bandwidth is small.

• The common parts of spectra are the all the more
  reduced that the spectral shift is critical.

17
The 2SIR simulator
2SIR is a simulator of radar images and
interferograms.
2. final radar image
summation for each range gate
Incident wave
4. Simulations with 2SIR.
It exploits hybrid techniques of ray-tracing and
z-buffering.
1. angular image
range
hidden echoes
tranparency
18
Examples
Radar image
Interferogram
Optical view
19
Simulations

• X band, VV polarisation.
• The sampling rate in range is equal to the
  resolution 0.5 meters.
• The interferometric system is bistatic.

• 2 types of case

Spatial Near range 800 km. Orthogonal
baseline 1...12 km
Airborne Near range 1.4 km. Orthogonal
baseline 2 m.
20
Example of reconstructed building
20 meters
mean profile
Profil
5. Results.
3D view
21
An other example of reconstruction
shape of reference
mean profile
With a different spectral shift
22
RMS as a function of the spectral shift
?f
Although the number of measurements is
insufficient to conclude, we can assume that
there is an optimal altitude of ambiguity.
23
Conclusions and prospects

• Unfolding layover area, slope filtering.

• Good results with simulations.

• Limitations

• Real data.

6. Conclusions.

• Double bounce.

• Quick analyser of slope.

24
Threshold V
It seems, we can modelize the behavior of the
threshold used on the vertical coherence. This
one decreases exponentially with the spectral
shift.
25
Double bounce
The strong echo due to the double bounce, reduces
the efficiency of this technique. A threshold on
the source images is required.
26
Vertical coherence on a slide
Coherence on a slope interferogram for a vertical
plane
Shape of reference
Kernel used to compute the coherence
27
Double bounce
Radar image
Interferogram
Optical view