**3.4.2 Result and analysis of collapsed building extraction**

Figure 7 shows the collapsed building distribution of the Yushu urban area extracted by the method. The rate of building collapse is about 58%. To verify the result, two test sites were selected as follows: (a) a severely damaged area, and (b) an almost undamaged area. Comparing the extraction results with the manual interpretation results from the airborne optical image with a resolution of 0.33 m shown as a' and b' in the lower left corner, we can see that the result of the collapsed buildings' extraction is consistent with the result from the optical manual interpretation. Furthermore, to verify the effectiveness of this method, the recognition rates of collapsed buildings and uncollapsed buildings are analyzed from two sample regions (more than 10000 pixels) from a and b test sites, respectively. From the statistical analysis, the recognition rate for collapsed buildings is 88% and that for uncollapsed buildings is 80%. It should be noted that the polarimetric SAR could play a more important role in the collapsed building extraction if the weather conditions were unsuitable for obtaining optical data.

Fig. 7. Collapsed buildings distribution of Yushu County extracted with the method from polarimetric SAR data, a and b in the lower left corner are the detailed images of the two test sites, a' and b' show the corresponding manual interpretation results from the airborne optical image(from Guo et al., 2010b).

, which can discriminate

method

, and the appropriate threshold value of

, the separation of collapsed and uncollapsed

<0.5 and

decomposition theorem, then, the bare soil surface was extracted with

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

Figure 7 shows the collapsed building distribution of the Yushu urban area extracted by the

 method. The rate of building collapse is about 58%. To verify the result, two test sites were selected as follows: (a) a severely damaged area, and (b) an almost undamaged area. Comparing the extraction results with the manual interpretation results from the airborne optical image with a resolution of 0.33 m shown as a' and b' in the lower left corner, we can see that the result of the collapsed buildings' extraction is consistent with the result from the optical manual interpretation. Furthermore, to verify the effectiveness of this method, the recognition rates of collapsed buildings and uncollapsed buildings are analyzed from two sample regions (more than 10000 pixels) from a and b test sites, respectively. From the statistical analysis, the recognition rate for collapsed buildings is 88% and that for uncollapsed buildings is 80%. It should be noted that the polarimetric SAR could play a more important role in the collapsed building extraction if the weather conditions were

buildings is conducted, and the distribution map of collapsed buildings is obtained.

Fig. 7. Collapsed buildings distribution of Yushu County extracted with the

airborne optical image(from Guo et al., 2010b).

from polarimetric SAR data, a and b in the lower left corner are the detailed images of the two test sites, a' and b' show the corresponding manual interpretation results from the

<42° (Cloude, 1996,1997). 2) Using statistical analysis, the

**3.4.2 Result and analysis of collapsed building extraction** 

image to analyze the statistical characteristics of

is obtained. 3) From the threshold value of

unsuitable for obtaining optical data.

the 

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 built of earth and wood.
