Transcript Document

Space remote sensing for urban damage
detection mapping and mitigation
Salvatore Stramondo1, Nazzareno Pierdicca2, Marco Chini3, Christian Bignami1
Istituto Nazionale di Geofisica e Vulcanologia,Via di Vigna Murata 605, 00143, Rome, Italy.
[email protected], [email protected]
2 Dept. of Electronic Engineering, Sapienza Univ. of Rome, Via Eudossiana 18, Rome, Italy.
[email protected]
3 Dept. of Physics, Univ. Alma Mater Studiorum of Bologna, Viale C. Berti Pichat 8, Bologna, Italy.
[email protected]
1
Introduction
In the last years the remote sensing techniques have been demonstrated a suitable monitoring tool for providing data useful for disaster mitigation.
In particular, in case of strong earthquake, the rapid detection of damaged buildings and infrastructures has assumed an important role for the civil
protection rescue activities. Moreover, the damage assessment can help the redevelopment process of the hit area. SAR has been revealed a
powerful instrument for change detection and damage evaluation purpose. In particular, interferometric features like the InSAR phase coherence and
the intensity correlation of multi-look images collected before and after an earthquake can be used to detect and quantify changes in built-up area.
On the other side, optical sensors have also been successfully used for damage estimation. In fact, the new optical sensors are reliable systems for
detecting changes of single buildings. However, the presence of clouds, shadows, variation in Sun illumination and geometric distortions are critical
for this type of sensors and prevent a fully automatic change detection approach. When both optical and SAR are available, a damage classification
can also be obtained by combining the two data types, leading to a more reliable result. This work proposes an effective procedure oriented to the
damage mapping. From data requirements (satellite images and auxiliary data) to product delivery the chain for damage mapping is described. This
latter in order to provide new instruments useful to Civil Protection Departments and Administrations for disaster management.
Data requirements
The minimum configuration to perform a damage
level estimation can be summarised as follow:
SAR data
 Two pre-seismic acquisition
 One post-seismic acquisition
OPICAL data
 One pre-seismic and one post-seismic
Auxiliary data
 local cartography
 DEM
The damage evaluation products can be also
generated by SAR data alone or by optical data alone
Optical Processing
EQ at time t0
OPTICAL
pre-seismic
OPTICAL
post-seismic
SAR-SLC t0+1
post-seismic
SAR-SLC t0-1
pre-seismic
SAR-SLC t0-2
pre-seismic
Data acquisition strategy
Data collection has to be done as
frequently as possible. In particular a
systematic approach of the acquisition
over high risk seismic area is preferable
This approach is intended to create a
catalogue, of both SAR and optical data,
composed by images as current as
possible, in order to avoid false alarm in
the damage detection caused by urban
development between pre-seismic and
post-seismic acquisitions
Pre-seismic complex coherence
Pre-seismic intensity correlation
Co-seismic complex coherence
Co-seismic intensity correlation
Change image
Auxiliary
data
Damage evaluation
approaches
The possible approaches for damage evaluation
can be separated in:
Dataset
composition:
co-registration
geo-coding
SAR processing
Change
image
SAR images are processed in order to
obtain two main SAR features: the
complex coherence and the intensity
correlation. The first one is calculated
from the single look complex data by
means of:

Damage
classification
procedure &
Layers
production
 Optical: when only optical data are available
change detection procedure give the map of
damages at different scale, depending on the
ground resolution of the sensors.
 Two images taken one pre- and
one post-seismic event.
 Co-registration of the images.
 Extraction of the other information
from
the
image
before
the
earthquake in order to have more
input for the building classification,
especially when we deal with
panchromatic images, just one
band.
 Building classification of the image
before the earthquake for limiting
the analysis of changes just at this
feature in order to reduce the
possible false alarms caused by
temporary objects (e.g. shadows,
cars, vegetation).
 Change detection procedure using
both images on the building map
which is based on the pre-seismic
image.
I 
 SAR & Optical: data fusion by classification
algorithms both parametric and non-parametric.
Product delivery
Description
Potential damage Prompt overview of potentially
snapshot
damaged area (only qualitative)

E s s 
E s1s1*
*
2 2
The second one is evaluated from multi
look data as follows:
 SAR: when only SAR data are available, the
difference between SAR features (complex
coherence and intensity correlation) is analysed
Product


E s1s2*
Tools
Scale
Medium/high res. OPT
& SAR
Medium, tens to
hundreds meters
Damage level at
district scale
Damage level (i.e., collapse ratio)
estimated on homogeneous urban
areas
medium/high res. OPT
& very-high res. OPT &
SAR
District, block of
buildings
Collapsed/heavy
damaged
buildings
Identification of single buildings
collapsed or heavily damaged
Very-high res. OPT
Single building
E ( I1I 2 )
E ( I12 ) E ( I 22 )
where s1 and s2 are the corresponding
complex pixel values, I1 and I2 are
corresponding pixel values of the
intensity, and E() indicates the expected
value.
Pre-seismic features are obtained
combining two pre-seismic acquisitions
whereas the co-seismic ones are derived
using one pre- and one post-event
acquisition