ANSI-C program: offset_pwr_trackingm.c
NAME
offset_pwr_trackingm - estimates range and azimuth offset fields
for detectec (MLI) images using intensity tracking
SYNOPSIS
offset_pwr_trackingm <pwr1> <pwr2>
<DIFF_par> <offs> <snr> [rwin] [azwin]
[offsets] [N_OVR] [thres] [rstep] [azstep] [rstart] [rstop]
[azstart] [azstop] [pflag]
| <pwr1> | (input) real valued intensity image 1 (reference) |
| <pwr2> | (input) real valued intensity image 2 |
| <DIFF_par> | DIFF/GEO parameter file |
| <offs> | (output) offset estimates (fcomplex) |
| <snr> | (output) offset estimation SNR (float) |
| [rwin] | search region size (range pixels, enter - for default from offset parameter file) |
| [azwin] | search region size (azimuth pixels, enter - for default from offset parameter file) |
| [offsets] | (output) range and azimuth offsets and SNR data in text format, enter - for no output |
| [N_OVR] | offset measurement oversampling factor (integer, default=1) |
| [thres] | offset estimation quality threshold (enter - for default=1.7) |
| [rstep] | step in range pixels (enter - for default: rwin/2) |
| [azstep] | step in azimuth pixels (enter - for default: azwin/2) |
| [rstart] | starting range pixel (enter - for default: rwin/2) |
| [rstop] | ending range pixel (enter - for default: nr - rwin/2) |
| [azstart] | starting azimuth line (enter - for default: azwin/2) |
| [azstop] | ending azimuth line (enter - for default: nlines - azwin/2) |
| [pflag] | print flag (default=0:print offset summary, 1:print all offset data) |
EXAMPLE
offset_pwr_trackingm 1352.mli 1610.mli 1352_1610.diff_par
1352_1610.offs 1352_1610.snr 32 32 1352_1610.offsets 2 7.0 10 10
400 1600 1200 2000
Estimates range and azimuth offset fields (offsets of 1610.mli relative to 1352.mli) using the speckle tracking algorithm using a 32 x 32 pixels search chip size and a 32 x 32 pixel search region size. An oversampling with a factor of 2 is applied. The search is repeated every 10 range and every 10 azimuth pixels, for the indicated reference image section.
DESCRIPTION
offset_pwr_trackingm estimates range and azimuth offset fields of a pair of MLI in support of image co-registration and offset tracking. The intensity tracking algorithm is used. Notice that this algorithm requires features in the detected images. The algorithm is not well suited for very small search chip sizes (< 16 x 16 pixels).
Throughout the image small MLI data chips are selected and used to determine the local range and azimuth offset fields. For each such offset estimation the intensity cross-correlation function is calculated. Oversampling of the correlation functon between the MLI images is supported. Normally this feature is turned off, however if precision measurement of the offset field is desired, increasing this value to 2, 3, or 4 will improve the accuracy. If the FFTW routines are not used, then the oversampling factor must itself be a power of 2. Values above 4 bring little or no benefit and will significantly slow the offset measurement. The location of the correlation maximum is estimated using a two-dimensional quadratic least squares fit.
The range and azimuth size of the rectangular chips used in the calculation of the correlation function corresponds to the search window sizes indicated. The search is centered at the offsets indicated in the offset polynomials of the provided offset parameter file.
The size of the search chips selected needs to be in-line with the features that shall be used in the matching and with the spatial dimension of expected variations in the offsets.
Offsets are calculated for the indicated section of the reference MLI (pwr1) at the indicated range and azimuth step. The related information is documented in the DIFF parameter file. For most cases a SNR threshold of 7.0 for the acceptance of an offset leads to good results. In cases of too low corelation the use of larger swearch chip and window sizes is recommended. Usually an oversampling rate of 2 leads already to very accurate offset estimates. The offset for the cross-correlation function is estimated using a two-dimensional least squares fit.
For a detailed investigation with many offsets estimated it is strongly recommended to provide offset polynomials in the DIFF parameter file. These polynomials are typically determined using the programs offset_pwrm and offset_fitm.
The estimated offset fields and related matching SNR values are written to binary output files of fcomplex (*.offs) and float (*.snr) format. The offsets can be converted to offsets in the original (reference) geometry using the program offset_trackingm. Or the program offset_fitm is used to cull the offset field and use the least squares derived polynomial fit for the offset between the two MLI images.
No search is conducted for chips with a NULL value in the center of the search chips. Setting the MLI values of areas which are not of interest to the NULL value may permit to strongly reduce the required procesing time, while still covering the area of interest at high spatial resolution (small range and azimuth step).
The binary offsets file (*.offs) contains offsets in SAR image pixels (of the reference image) in fcomplex format. The real parts correspond to range offsets and the imaginary parts to azimuth offsets. The positions of the offset estimates are defined by the starting range and azimuth positions and the range and azimuth steps, as documented in the DIFF parameter file.
OPTIONS
Range and azimuth chip sizes / serach region sizes, and range and
azimuth steps can all be indicated on the command line,
permitting for example to use search chips, search windows, and
steps which are different in rnage and azimuth directions.
SEE ALSO
typedef_ISP.h, typedef_DIFF.h, DIFF_par, offset_trackingm, offset_fitm, offset_pwrm, create_diff_par. .
DIAGNOSTICS
All messages are generally self-explanatory.
NOTES
None.