**3.4 Extraction and analysis of collapsed buildings from polarimetric SAR**

Large-scale earthquakes severely damage people's lives and property. Fast, accurate, and effective collapsed buildings a monitoring and evaluation after earthquake using remote sensing provides an important scientific basis and decision-making support for government emergency command and post-disaster reconstruction.

RADARSAT-2 polarimetric SAR data (FQ mode, ascending) on April 21, 2010 from Yushu County were used to extract the distribution of collapsed buildings. The resolution of the image is about 8 m and the incidence angle is 21°. From this polarimetric SAR data, the H method (Guo et al., 2010b) was used to extract the spatial distribution of building collapse caused by the earthquake in the Yushu urban area. At the same time, for comparison and analysis, a manual interpretation map obtained from the high-resolution airborne optical image was also collected and shown in Figure 6. From the collapsed buildings extraction result, the reasons for the severe earthquake damage to buildings are also discussed and assessed.

Fig. 6. Spatial distribution of collapsed buildings interpreted from airborne images and sample images for typical regions.

#### **3.4.1** *H***method**

304 Earthquake Research and Analysis – Statistical Studies, Observations and Planning

each side of the fault in the low coherence area, and the cross-fault displacement in the line

Fig. 5. Coseismic deformation map from ALOS PALSAR data, (a) Differential

**3.4 Extraction and analysis of collapsed buildings from polarimetric SAR** 

emergency command and post-disaster reconstruction.

interferometric phase map; (b) Differential interferometric phase of instrumental epicenter; (c) differential interferometric phase of macro epicenter. A, B, C1, C2, D1 and D2 in (a) represent the different positions and in (b) and in (c) are two large deformation areas (from

Large-scale earthquakes severely damage people's lives and property. Fast, accurate, and effective collapsed buildings a monitoring and evaluation after earthquake using remote sensing provides an important scientific basis and decision-making support for government

RADARSAT-2 polarimetric SAR data (FQ mode, ascending) on April 21, 2010 from Yushu County were used to extract the distribution of collapsed buildings. The resolution of the image is about 8 m and the incidence angle is 21°. From this polarimetric SAR data, the H method (Guo et al., 2010b) was used to extract the spatial distribution of building collapse caused by the earthquake in the Yushu urban area. At the same time, for comparison and analysis, a manual interpretation map obtained from the high-resolution airborne optical image was also collected and shown in Figure 6. From the collapsed buildings extraction result, the reasons for the severe earthquake damage to buildings are

of sight should not be less than 11.8×8=94.4 cm.

Guo et al., 2010b).

also discussed and assessed.

From the RADARSAT-2 polarimetric data (FQ mode) and the polarimetric decomposition model, a new method that uses only one post-earthquake SAR image was proposed to identify collapsed buildings. This method mainly utilizes three important polarimetric parameters to extract the collapsed buildings. These parameters are *H*, and , Where is entropy, representing the random level of target scattering, is the averaged scattering type and is the circle polarization correlation coefficient which is very sensitive to artificial objects. and are obtained by using Cloude's decomposition(Cloude, 1996,1997).The of uncollapsed buildings is high while that of collapsed buildings is low. However, is also related to the surface roughness. For a low roughness surface, the is also high. Therefore, it is necessary to remove the disadvantaged influence of the bare soil surface before this parameter can be used for building identification( Ainsworth etc al.,2008 ). The circular polarization correlation coefficient ρ can be expressed as Eq. 6( F.Mattia, 1997),

$$\rho\_{RRLL} = \frac{}{\sqrt{<|S\_{RR}|^2><|S\_{LL}|^2>}}\tag{6}$$

Where ��� = ���� <sup>+</sup> � � ���� − ���), ��� = ���� <sup>−</sup> � � ���� − ���). For the Yushu urban area, three main land cover types were categorized: collapsed buildings, uncollapsed buildings, and bare soil surface. The basic process using method to identify the collapsed buildings is as follows: 1) the extraction of the bare soil surface. and are obtained using

Earth Observation for Earthquake Disaster Monitoring and Assessment 307

From the collapsed building distribution, it is clear that the degree of collapse is related to the distance from the main fault. The buildings, which were constructed in the alluvial zone, had very poor earthquake resistance because their foundations were weak and most were

Based on remote sensing image two-dimensional spatial information systems, interpretation and analysis has been the main means of seismic disaster assessment. Although twodimensional spatial information systems have the macroscopic and overall characteristics, it also has inevitable defects in the earthquake disaster assessment (Xiao et al, 2001, Li et al, 2007), such as in accurate expression of three-dimensional information, an inability to record non-uniform three-dimensional spatial entities, and a lack of a basis for uneven spatial entity description of lake water, landslides, collapsed buildings and so on. Therefore, it is necessary to establish an evaluation system based on three-dimensional spatial information

For the Wenchuan and Yushu earthquakes, using three-dimensional simulation and evaluation with advanced Earth observation technology, we established realistic 3D terrain model using the relatively sophisticated disaster assessment model and successfully created an earthquake disaster simulation and evaluation system. Based on multi-sensor, multitemporal, and multi-resolution remote sensing images and 1: 50,000 scale DEM data, we produced technology for large 3D terrain modeling and interactive real-time rendering. The 3D simulation system provides a more intuitive disaster analysis method for major collapse, landslide, and debris flow disasters. Using red-blue 3D imaging techniques to get airborne remote sensing stereo images and reconstruct three-dimensional scenes, we can efficiently extract data on damage to housing and improve the disaster analysis accuracy using red and blue stereo glasses. In a 3D environment, not only can the geo-spatial relationship between objects be shown, but also the topological relations between spatial objects. Practice has

proven that this kind of three-dimensional assessment is more efficient and reliable.

We use 1: 50,000 topographic vector data of the disaster area for error analysis and to eliminate gross errors of contour and control point data. We use a difference algorithm for vector contour data to obtain high-resolution DEM raster data. In order to rebuild 3D virtual scenes, we use a merging method of aerial remote sensing images and Landsat TM images to generate the terrain texture. In severe disaster areas, we acquired highresolution data, TM images. Then combined with DEM data, the aerial remote sensing data can be corrected precisely. Finally, the merging of aviation data and TM data yield the an image of the entire area, which is then mapped to the three-dimensional terrain

The complicated terrain of earthquake-stricken areas and large data of three-dimension model after overlaying images and DEM have brought challenges for real-time rendering. This paper proposes a multi-resolution triangular grid dynamic geographic model based on computing vision, established a simplified algorithm based on multi-resolution vision. Using a multi-resolution scene model algorithm, this paper resolves the real-time interactive roaming difficulty of large-scale three-dimensional terrain data. First, according to the

**4.1 Three-dimensional terrain modeling and visualization** 

model to form a virtual 3D environment.

**4. The foundation of an earthquake condition simulation and evaluation** 

built of earth and wood.

of earthquake disaster information for analysis.

**system** 

the decomposition theorem, then, the bare soil surface was extracted with <0.5 and <42° (Cloude, 1996,1997). 2) Using statistical analysis, the , which can discriminate between uncollapsed and collapsed buildings, is determined. From the high- resolution optical data, typical areas of collapsed and uncollapsed buildings are selected from the radar image to analyze the statistical characteristics of , and the appropriate threshold value of is obtained. 3) From the threshold value of , the separation of collapsed and uncollapsed buildings is conducted, and the distribution map of collapsed buildings is obtained.
