3. Data Processing Procedure
The orbit and the observing angle of the ERS satellites are known to be fixed. Hence, it is well suitable for the Interferometry measurements. In another words, it is quite a stable dual pass or repeat pass system. We processed the ERS2 raw signals of these four images (21 Jan,6 May,23 Sep and 28 Oct ) with single look complex(SLC) product. Their coherence maps are respectively displayed in Figure.2 to Figure.5. Noted that the signal data being processed are in complex form, with 16-bit representations for each of the real and imaginary parts, so we can preserve the original phase information. The pixel spacing in slant range and azimuth are 7.9 meters and 3.9 meters, respectively. First we processed the pair of 21 Jan and 6 May before earthquake, and 23 Sep and 28 Oct after earthquake. In view of these two pairs of interference fringes, it can be clearly observed that they are correlated to each other, and there is little atmosphere inference. The differential interference fringes implies the presence of the surface deformation by earthquake. It is noted that the spacing of the fringes is obviously in connection with the base line; great base line has great spacing and vise visa.
We proceed in the way called"Three Pass"by using image from May 6 as reference scene before earthquake, and matching that with 23 Sep and 28 Oct to produce two interferogram images removing the earth curved surface and topography inference. At this point, we still keep working in slant rage coordinate, so that we can subtract with the interferogram images (May 6 and Jan 21) directly to generate two interferometric phase images due to surface deformation, (6 May, 23 Sep) and (6 May, 28 Oct) (Figure.6 and Figure.7 ). Finally, we perform phase restoration in tenth of pixel accuracy to get the surface deformation.
The next step is the image pairs matching each other by conjugate points. Since the orbits are nearly parallel, this kind of images are similar to each other in the geometric characteristic. The conjugate points is best uniformly distributed over the study area and the required number depends on the terrain relief. For this study, about 100 were used. The rms error registration is about 0.1 pixel. Then, the single look complex data is resampled, followed by a 2x10 running window to reduce the speckle noise. The method phase unwrapping is then applied to produce the image of the change of the topography. In this way, the phase error of the interferogram caused by coregistration might be minimized.
Figure2: The Coherence map on May 6 and Jan 21
Figure3: The Coherence map on May 6 and Sep 23
Figure. 4: The Coherence map on May 6 and Oct 28
Figure. 5: The Coherence map on Sep 23 and Oct 28
Figure. 6: The differential interferogram on May 6 and Sep 23
Figure. 7: The differential interferogram
on May 6 and Oct 28
