Effect of coherence on dems derived from sar interferometry:
A case study of mayon volcano, philippines
Vu Tuong Thuy and Mitsuharu Tokunaga
Space Technology Applications and Research Program
Asian Institute of Technology
P.O. Box 4 Klong Luang, Pathumthani, 12120, Thailand
Tel: +66-2-5245577 or +66-2-5245690 Fax: +66-2-5245597
Email: rsc009994@ait.ac.th , tokunaga@ait.ac.th
THAILAND
Keywords : SAR Interferometry, DEM, phase coherence
ABSTRACT :
Synthetic Aperture Radar (SAR) Interferometry is a technique for obtaining the Digital Elevation Model (DEM) of the Earth surface by processing the phase difference between coincidence SAR images. Nowadays, SAR interferometry is considered as the promising technique for obtaining high resolution and quickly updated DEM. Here we present two experiments that make use of ERS1&2 Tandem data for making DEM derived from SAR Inteferometry. As the results, we show the limitation of the length of baseline for C band data, the correlation of DEM accuracy with coherence, and the correlation of DEM accuracy with NDVI.
1. Introduction
Recent years, Synthetic Aperture Radar (SAR) Interferometry has been utilized as a potential technique for obtaining accurate DEM of the Earth surface. The SAR interferometry products are becoming familiar to the applications and researches. SAR interferometry is based on the processing the phase difference between coincident SAR images. The displacement vector between two antennas is called baseline vector. Two antennas may be mounted on one aircraft and form along-track interferometry or across-track interferometry. The other way utilized in this research is repeat-pass interferometry, which is the normal implementation for satellite. If the ground objects still keep the high correlation between two times of illumination, the spatial baseline can be synthesized.
There have been several researches (Wehr A., et al., 1996, and Zebker H.A., et al., 1994) pointing out some results of the effect of phase coherence on DEM accuracy as well as analysis and utilizing of phase coherence. However, there is no any this kind of result for Southeast Asian tropical region. In this paper, besides the utilizing ERS1&2 data for investigating effect of phase coherence on DEM of Mayon volcano, we investigated the relationship between phase coherence and land cover.
2. Methodology
Table 1 - Data set used
| Data |
Date |
Characteristics |
| ERS 1 & 2 Tandem |
20 and 21 May 1996 |
Ascending |
| ERS 1 & 2Tandem |
10 and 11 October 1997 |
Descending |
| Landsat TM |
18 February 1994 |
Geocoded |
| DEM |
Topographic map |
Geocoded |
In general, the processing can be separated into three parts as below:
2.1 Making DEM
Calculation of interferograms and coherence coefficient
First, two single-complex images were coregistrated together. After coregistrating two images and resampling one to other, the complex inteferogram was calculated by using the following equation:
where s1 and s2 are single look complex values of two coregistered images,
* stands for conjugate complex,
<> stands for ensemble average
A portion of interferometric phase that are affected by the orbital fringe should be removed out of interferogram based on perpendicular component of baseline, then flattened inteferogram was formed. Along with this processing, coherence coefficient also was estimated.
Phase unwrapping
Before doing phase unwrapping, the interferogram was filtered to reduce phase noise. Then, phase unwrapping was done following region-growing method. The low correlation area, layover area and neutral trees that were formed by the connection of residues were masked to avoid in phase unwrapping.
Coregistration between DEM derived from topographic map and SAR slant-range
Here, GCPs were extracted from DEM derived from topographic map (DEMtopo) to precisely estimate baseline geometry, so it is necessary to coregistration between DEMtopo and SAR image. The automatically registration method is based on the cross-correlation between SAR image and simulated SAR image from DEMtopo.
Creation of slant-range and ground-range elevation images
Based on the extracted GCPs above, least squared method was used to precisely estimate baseline geometry.
Now, we can calculate the elevetion from unwrapped phase
The equations used are listed as follow:
B, Bc, Bn, r1 and r2 (see Figure 1) are known values.
