This page describes the status of the Doris software.
The project for the development of the Doris software
started September 1st, 1998.
After one year we decided to make the software
freely available to the scientific community.
We hope the further development of Doris will be a joint effort.
The first version of Doris was capable of generating
the classic interferometric products starting from SLC images.
This release included user's documentation.
The next stage in the development was to create a software package that
could do all procesing from RAW data to products
such as (geocoded) deformation maps.
Doris uses SLC images as input however, and will remain to do so.
In the approach used at Delft University of Technology, we
can create SLC images from RAW data for various sensors,
and then process the SLC images to interferometric products using Doris.
Unfortunately, the SAR processor
(for focussing the RAW data, i.e., to create an SLC)
cannot be placed in the public domain by us.
Currently, Doris can handle SLC data from the satellites
ERS, JERS, ENVISAT and RADARSAT-1/2, ALOS, TSX and COSMO-SkyMed.
With the release of snaphu unwrapping software in the public domain by Curtis Cheng, SU, an important processing step could be included in Doris v3.4.
Visualization and data interpretation is generally
performed by the cpxfiddle utility and/or GMT software,
based on the interferometric products created by Doris.
Doris can also automatically create quicklook figures
in SUNraster format.
1) The script doris.process-reset.sh is renamed to doris.rmstep.sh and is updated for Mac OS X. 2) plotcpm is updated to plot offsets rather than the offset residuals. Another script is made available under the name plotcpm.residues to plot the offset residuals. Bug fixes: 3) Step COMPREFDEM: when multilooking applied during computation of reference DEM phase, the output file was smaller than the expected size. 4) Step COMPREFPHA: fixed index overflow in sinus look up table. Please see the manual for details.
1) Handling of file sizes larger than 4GB on both 32-bit and 64-bit platforms. 2) Processing for TerraSAR-X data. 3) Improvement of the master-slave overlap calculation. The new algorithm should prevent segmentation faults as sometimes experienced in the past. 4) Improvement and speed-up of COMPREFDEM based on a two-step approach. First, the complete DEM is radarcoded and written to a file. Second, the radarcoded DEM is interpolated to the master image geometry using Delaunay triangulation. The overlapping buffers, as in previous versions of Doris, are prevented. 5) Simulation of master amplitude image based on DEM (M_SIMAMP). 6) Estimation of master timing error based on correlation between master amplitude and simulated master amplitude (M_TIMING). 7) Estimation of relative timing error between master and slave image (RELTIMING). 8) DEM assisted coregistration (DEMASSIST). For optimal performance, apply after M_SIMAMP, M_TIMING and RELTIMING. 9) Coherence estimation after DEM subtraction (default). 10) Utility construct_dem.sh to download, merge and fill voids of SRTM data. 11) Utility doris.process-reset.sh to reset and clean up the processing entries in Doris result files allowing to easily re-run a step or multiple steps. 12) Option to output height-to-phase conversion factors (H2PH) based on flat earth (SUBTRREFPHA) or DEM (COMPREFDEM). 13) During COARSECORR and FINE coregistration NaN (not-a-number) correlation values, which are mainly due to no data regions at the border of the scene, are eliminated automatically, but the whole list of windows are kept in the doris.log file. 14) Coherence output is properly scaled between 0 to 1 range with new option -r of cpxfiddle. Please see the manual for details.
1) Several bug fixes + new compiler supports. 2) Experimental support to ALOS 3) Generating configuration file to use multiple CPUs during unwrapping (Snaphu) 4) Now RS_DBOW_GE0 is available during resampling step 5) Updates to default values of some parameters. Please see the source code for details.
1) Added heightamb utility to bin dir (try it). 2) Added raised cosine interpolation kernel for interpolation: (new argument "RC_6" and "RC_12" for card RS_KERNEL); See paper cho05; RC_6 is now the default kernel in run. 3) Getoffset more stable to figure out the offset for COARSE_CORR: The estimates are now sorted on the estimated correlation and The mean offset of the highest correlation windows is used. 4) Slc info class: more robust to look for parameters in .res file: Removed updateslcimageML function (integrated with updateslcimage); 5) Added azimuth oversampling in PROCESS OVERSAMPLE: Factor 2 in azimuth now supported for PS type analysis. 6) Added method OVERSAMPLE in process step FINE: First oversamples the complex data by a factor two before cross-correlating the magnitude patches. 7) Changed behavior of FINE: first a local peak is found in window of size FC_WINSIZE then peak is refined. e.g., if winsize=64, then correlations for offsets between -32 and 32 are estimated; then around peak correlation a chip of size ACC is cropped, which is oversampled. cross-correlation of zero-meaned magnitude is performed. 8) Added method CPM_WEIGHT BAMLER (this is the new default method): See papers IGARSS 2000, 2004 by R.Bamler. The standard deviation of the estimated shift is estimated and used to estimate the coregistration polynomial. This may generate some new warnings, but I believe it to be a better method for weighting the observations (estimated offsets). 9) LAPACK: Changed default to using fortran version of Lapack library. This should make it possible to use LAPACK for most Linux systems. 10) Added -DWIN32 for some people that want to compile doris directly under windows. This is not supported and not adviced, but I wanted to avoid diverging code trees (I will not answer questions about running Doris directly under windows). Most likely a few more changes are required, regarding using the name "DEBUG" and "min" which conflict Changes in "run" templates: RS_KERNEL RC6 // new kernel, optimal for oversampled data. CC_WINSIZE 256 256 // was 512 128 but suspect strange effect if not square FC_METHOD oversample // experimental, best, but slowest CPM_WEIGHT bamler // experimental, but expected to be best Some smaller issues are also resolved, e.g., more verbose output and some reported problems with exception handling. It seems that a SIGSEGV error still occurs occasionally at some Linux systems. Since I do not experience these errors I cannot debug it. If you get these errors, try to compile without the -O flag (edit the Makefile). But first, as always when you get an error, compile a debug version of Doris (make clean, configure, make) and repeat the processing with that version using SCREEN trace. See the file "newsincev3.4" and the source code for details. Also see the Doxygen documentation of the source on website. http://doris.tudelft.nl/DOXYGEN/html/ Doris3.16 is compiled and tested at SUN Solaris using gcc 3.3.1 and 2.95.1, and on Linux systems (I don't know the compiler version.) Best regards, Bert Kampes
Mainly TRILINEAR method and OUT_FILE_DEM_LP for COMPREFDEM;
Atlantis processed ERS data reader;
exceptions.cc
BASELINE class
1) Small bug in fDC computation during resampling fixed. This bug had no practical consequences for ERS processing. 2) Small bug in configure script fixed for detection of Endian type. 3) Changed source code to .hh include files; Changed some conversion routines to be members of class slcimage.
1) coherence:
. added check for zero division that caused snaphu
to exit due to NaNs in the coherence file
. removed some lines that wrote zero lines at start;
I am not sure why those lines where there, although likely
they fixed some situation. In the run file, the window
for coherence computation is set to 32x32, for multilooking
to 10x2 (to reduce the bias of the estimated coherence).
2) adaptive range filter:
. now uses last know spectral shift if a peak cannot be detected.
. removed debug code causing large output files.
3) goldstein phase filter:
. I think there was a bug in the code here when a smoothing
kernel was specified. (fixed)
-RADARSAT processing (CEOS reader also for Atlantis SAR processor). -Bugfix related to ENVISAT processing when data was cropped for slave image.
-New cards HEIGHT, TIEPOINT, AZ_TIMING_ERROR, RG_TIMING_ERROR -Bugfix envisat tools for cropping of area and more robust reading of meta information.
-included geographic cropping. -included oversampling in range of cropped data (thanks to R.Nutricato). -JERS processing.
-included fftw library, speed up upto 50 percent. -quicklook option processing for run script -clarity of defines and configure script.
-Automated coregistration polynomial computation. (cards CPM_MAXITER, CPM_KALPHA, CPM_WEIGHT QUADRATIC) -Software prepared for JERS, RadarSAT data formats. -Coherence computation now checks if it divides by zero. -Added scripts in bin dir to show how to create DEMs like this. -Minor bugfixes in cpxfiddle, scripts to make it more generic.
-This version can handle ENVISAT SLC data. argument M_RF_METHOD ASAR now works correctly. -Orbit interpolation default method changed to polynomial fit. card POLYFIT -Minor bugfixes in scripts and the software. output format of azimuth filter if option CI2 selected.
...
...
updatefile("scratchlogdat",input_general.logfile);
break; //Raffaele Nutricato reported missing break here!
// ______ ENVISAT ASAR ______
case readfiles_asar:
ERROR("working on this, use system call hack?");
break;
default:
ERROR("impossible error for checked input.");
}
BEEP (level)
PREVIEW (create quicklooks sunraster files with cpxfiddle program,
to quickly evaluate results. Download and install
separately from SARtools.tar.gz)
UNWRAP (wrapper for snaphu program, download and install separately.)
This completes/makes uniform an important step missing so far
in Doris.
Improvements have been made to steps:
COMPREFDEM (slower, but more robust)
RESAMPLE (azimuth kernel is shifted to Doppler centroid, default
behavior.)
Improvements have been made to source code organization:
Renamed source code files .c --> .cc
Changed include files to e.g., -->
run script some small bugs.
Hopefully this will make Doris easier to compile with the new
gcc compiler. And that it does not affect other users with older
compilers. It has been tested only under cygwin so far, gcc version
2.95.3-5 (cygwin special).
|
This page is maintained by TUDelft - MGP Radar Group, mail to: doris_users#at#tudelft.nl |
|
Disclaimer: These pages contain links to sites outside DUT. We do not take responsibility for the contents of those sites. |