Characteristics of Satellite SAR Images in the Damaged Areas Due to the Hyogoken-Nanbu Earthquake
Calculated Characteristics of Processing Images
The correlation image obtained by the procedure above is shown in Fig. 3 black color indicates areas where the correlation coefficient is low. Comparing the urban area around Osaka and the forest areas, the correlation coefficient of the urban area is higher that of the forest area. The difference before and after the earthquake is considered to reflect the change of land cover. The correlation in the urban area from Kobe to Nishinomiya is slightly lower than other urban areas. The difference image is shown in Fig. 4. black color indicated area where the difference of the backscattering coefficients is significant. A black zone is seen conpicuosuly in the urban area stretching from kobe to Nishinomiya.
Fig.3 Correlation of the intensity images before and after the Hyogoken-nanbu The Hyogoken-nanbu earthquake
Fig. 4 Difference of backscattering cofficient Obtained from SAR images before and after the Hyogoken-nanbu earthquake
Characteristics of Satellite SAR Images in the Damaged Area
The correlation image and the difference image are compared with the damage ratio of buildings evaluated from the field survey data. The severe damage ratio is defined as the ratio of the number of seriously damaged buildings to the total number of buildings within a district block. The relationship between the average difference of the backscattering coefficients and the average correlation coefficient for different severe damage ratio, D, is shown in Fig. 5 the mean values and the standard deviations by the correlation analysis are listed in Table 2.
Fig.5 Relationship between the average backscattering difference and the average correlation coefficient
| |
Correlation
coefficient |
Backscattering
difference [dB] |
| Server damage ratio D(%) |
Average m |
Standard Deviations |
Averagem |
Standard Deviation s |
| No damage |
0.32 |
0.27 |
0.59 |
1.38 |
| D=0 |
0.30 |
0.26 |
0.38 |
1.40 |
| D£30 |
0.25 |
0.26 |
0.12 |
1.44 |
| D£60 |
0.20 |
0.27 |
-0.40 |
1.72 |
| 60<D |
0.16 |
0.27 |
-0.75 |
1.96 |
Table 2 Average correlation coefficient and backscattering difference for the classified damaged area.
Nodamaged areas indicate the pixels where the houses have no exterior evidence of damage due to the field survey. The results show that the difference of the backscattering coefficients becomes high and negative and the correlation coefficient becomes low in the area of a high severe damage. On the other hand, in the area of a low severe damage. On the other hand, in the area of a low severe damage ratio, the difference of the backscattering coefficients becomes high and positive and the correlation coefficient becomes high. However, since the standard deviations of the above two values are quite large, a large amount of randomness is included in the damage identification based on the method using the two values.
Generally, artificial structure show comparatively large reflection because of the specular characteristics of the structure and the ground. Open space has comparatively small reflection because of its smooth surface. Following the earthquake, structures were destroyed and removed leaving the plain exposed. Therefore, it is considered that the backscattering intensity taken after the earthquake is surface exerts an influence on the decrease of the correlation. It is desirable that the difference of the backscattering coefficients of the nondamaged area become zero. However, the results were somewhat different due to the following reasons:
-
The observation conditions at the two time instants are not the same. For example, the satellite orbit is not necessarily equal for the two instants.
-
- Even in a nondamaged area, there is a change in earth's surface whether covered by building or vegetation.
- There is slight influence from the atmosphere.
- Accuracy of geometric corrected images
