Error in coregistration step with ALOS data

This topic contains 1 reply, has 2 voices, and was last updated by  periz 3 years, 10 months ago.

  • Author
  • #723


    Dear Daniele,

    I started to process 2 ALOS data, but after coregistration step i get this error:

    Error in cull_points(): fewer than minimum points, only 17 left!
    WARNING!!! coregistration of image 20101212 failed!!!!

    And then i try to start again (re-run the tool, using SLC data module), “get contents”, check “update”, set orbits, then go to coreg params,check initial shifts from coarse xcorr (my previous procesiing i didn’t change anything from coreg params), then coregister again. Finally i get this warning:

    param2orbitstruct: reading from SLC data
    Task for UTC time: 08-Mar-2009 20:32:00 and LOCATION: (37.742, 28.835) finished.

    Slave images coregistration

    get_master_pts: using the previously generated file.

    To re-generate it, remove file /home/fatma/Desktop/babadag_sarproz/babadag/EXT/master_pts_infile.txt

    Using a coarse cross-correlation to calculate initial co-registration shifts

    cross-correlating matrixes

    Image pair: 20090308-20101212
    The first 5 local maxima have the following reliabilities:
    2.1163 0.1047 0.0991 0.0979 0.0953

    Estimated shifts: 80, 40

    Xcorr correction: -61.0 in range, -5.0 in azimuth

    allocating 4300 x 10560 samples, type: 1

    allocating 4300 x 10560 samples, type: 3

    prep_img: job ended

    setimageconf: converting from SLC data

    image.conf.mat written for 1 images

    set3points: converting from SLC data

    orbit files written for 1 images

    param2orbitstruct: reading from SLC data
    Task for UTC time: 12-Dec-2010 20:32:00 and LOCATION: (37.742, 28.832) finished.


    pair 20090308_20101212: highly probably the shift is too large and the co-registered image has black areas!!!!

    The Processing concluded successfully, here a specific message:

    Slave images coregistration: job ended

    Elapsed Time: 883 seconds
    File 20101212_HH.slc not found
    1 Slave images have not been extracted!!

    And also you can see screenshot from my procesiing steps. How can i solve this problem?

    Thank you,


    You must be logged in to view attached files.
  • #725


    The software has some default parameters for coregistration.
    If the coregistration fails at the first step, one has to plot the images to check why it failed.
    Then, depending on the reason, you can change the coregistration parameters.

    In your case, the two images look similar and have corresponding features that can be used for coregistering them.
    So, the reason of the failure is that the default parameters (12 “search pixels” in range/azimuth) are not enough to compensate for the shift between Master and Slave. It means that the timing info coming with the orbits is not accurate enough.
    This can be solved by modifying the initial shifts: instead of using the orbits, we can use a coarse cross-correlation (tick the corresponding checkbox). (note: you could also increase the “search pixels” values, but this will take longer processing time)

    The second message that you got tells you that the estimated shift between the two images is higher than the “buffer” taken for the slave image. It means that you could have black pixels at the border of you output slave.
    To solve it, one should increase the value of “delta” in the “data processing” frame, “SLC processing” window. After changing that value, check the box “Ovw” (overwrite) and extract the slave image again. Finally, launch the coregistration.
    However, it may not be necessary: probably if you plot the co-registered slave you will not even notice the black border.
    If this is the case, you can proceed further with your processing.

    Notice that if the coregistration (or any other steps) fails, generally you do not need to repeat processing steps that were already concluded. Make use of the “update”, “overwrite” options and ALWAYS read the messages given by the software.

You must be logged in to reply to this topic.