137
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
temporary stand-points, because the replacement of the
instrument hasn’t been foreseen for this test. As shown in figure
6, the radar system has been placed just in front of the dam’s
downstream face. This positioning has allowed to capture the
entire dam displacement along in the range direction during the
whole observation time. The time needed to scan the whole
downstream has been 9 minutes, and each scanning cycle has
been repeated without intermediate breaks. It is relevant to
notice that the configuration of parameters needed to correctly
perform the data acquisition is very easy, because this task can
be performed by also non-skilled operators. During the data
acquisition, the control software allows one to check the
focalisation map, which is an information of the reflectivity
properties of the object illuminated by the SAR sensor. In
addition, a check on intermediate displacements of points
tracked on the structure can be seen. This option is very
important for continuous monitoring applications.
*
Fig. 6 - IBIS-L positioned in front of the downstream face of
the Cancano dam (Alta Valtellina, Italy)
The results of GBInSAR measurements are 2-D deformation
maps of displacements, which reached a maximum value of
about 4 mm in the central section. Consider that, in years with
the biggest variation of water level in the basin, the full
displacement of this point might reach about 80 mm. In figure 7,
maps corresponding to three different epochs (after 10, 22 and
37 h) during the test at Cancano dam are shown.
According to an integer ambiguity of 4.5 mm (Acp= 1 rad), such
a displacement can be automatically tracked if IBIS-L is
continuosly acquiring data, and not more that 1 cycle is
increased from an observation epoch and another (in this
application these are separated by a 9 minute rate). In case of
repositioning of IBIS-L, the processing tool is not able to detect
displacements larger than the integer ambiguity. A solution, to
this problem could be provided by integrating the GBInSAR
system to other ranging sensors featuring lower precision and
measurement rate, but that would be able to reconstruct low
frequency displacements (e.g. TLS or robotic total stations).
By thresholding points on the basis of their coherence (p) and
their Signal-to-Noise Ratio (SNR), it is possible to select those
which are high accurate control points. The higher are the
selected thresholds, the higher is the accuracy in the
measurement of the displacement of the selected point. In
figure 8, a sub-set S0.99 of only points with p>0.99 is shown. For
each of these the time-series of displacements is fully available.
For 5 points on the dam crest, the displacements are shown in
the upper part of figure 9, while in the lower part displacements
of points which are vertically widespread are shown. As you
can see, the trend of the observed dam displacements agrees to
the foreseen static behavior in both cases. Note the magnitude
of maxima displacements along the observation period is under
2 mm. Some outliers are present in time-series corresponding to
points P34 and P30, which probably are placed on positions
disturbed by local turbulence or they might be located on a
surface featuring bad reflectivity, despite of the high coherence.
On the other hand, these outliers does not result in trend errors,
and can be removed by low-pass filtering in the time domain, or
by averaging displacements evaluated on close points in the
space domain. Indeed, by comparing displacements of points
P30 and P29, whose position differ for only 15 m, it is possible
to see that only the first of them is affected by anomalous noise
(see Fig. 9).
Fig. 7 - Complete displacement maps after 10, 22 and 37
hours of observation
Fig. 8 - Location of tracked points on the dam downstream face
with a coherence p>0.99