A SAR satellite does not downlink images. To turn raw satellite signal data into an image, you must use a SAR processor. Satellite imaging organizations such as ASF or RSI usually provide users with processed SAR images.
For interferometry, however, you can achieve better results by doing your own SAR processing. For example, if you process both images to an averaged doppler, you minimize rotational decorrelation and maximize the interferogram coherence.
For this reason, you may wish to use AISP, the ASF-STEP Lab's SAR processor. AISP works well with ASF CCSD data, but can be easily adapted to work with any kind of signal data. AISP has sucessfully processed images from ERS-1, ERS-2, JERS, and all the beams of RADARSAT (fine, standard, extended, and ScanSAR).
For a SAR processor, AISP is fast. On a single processor, a full ERS frame takes about 50 minutes to process. A parallel implementation of AISP, PIASP, can process a full ERS frame in under 2 minutes (using 56 processors on a Cray T3E).
AISP can be run straight from the command line, but it is usually run for interferometry from the Register_ccsd script. This script takes care of averaging the doppler, resampling the second image (as it is processed), and creating an interferogram.
You can also run AISP alone. It will, by default, generate both a full-resolution complex as well as a multilooked amplitude image.
AISP works on a "patch" of data at a time. A patch is a chunk of SAR data small enough to fit all in memory, but large enough to efficiently compress.
AISP begins work on a patch by reading in and unpacking the data for the patch. It then range-compresses the patch by correlating the data with a range pulse replica. This replica can either be synthesized by AISP (linear FM); or read in from a file.
The data is then transposed and FFT'd along azimuth. The patch is then range migrated to straighten the lines along azimuth. Finally, the processor constructs an azimuth doppler history replica for each line, and azimuth compresses each line. The processor includes the ability to resample the image as it is range migrated and azimuth compressed; this saves a resampling step later on.
The final product is a fully phase-preserved, floating-point single-look complex image (.cpx). A multilooked floating-point amplitude image is also generated (_amp.img).
Back to main interferometry page.
Last Updated: September 1, 1998
If you have any questions, please feel free to email olawlor@acm.org