sarproz-group

Slight offset after orthorectification (S1)

This topic contains 6 replies, has 2 voices, and was last updated by  steffan 8 months, 2 weeks ago.

  • Author
    Posts
  • #2879

    steffan
    Participant

    Hello all,

    I am currently working with 60 Sentinel 1 images taken in 2017. I am processing change detection over a very large area. Orthorectification does a great job but there is still a slight shift between SAR results and Google Earth, in some places as much as 100+ meters. This is especially noticeable over lakes or center pivot irrigation fields (example annexed).

    Is there any way to improve the overlay?

    Thank you.

    • This topic was modified 8 months, 3 weeks ago by  steffan. Reason: Added another example
    Attachments:
    You must be logged in to view attached files.
  • #2884

    periz
    Keymaster

    Hi Steffan,
    to properly judge the geo-referencing you should observe carefully the reflectivity map (not the changes) and you should look at multiple areas. If all areas present the same coherent shift, then it is a problem that can be corrected by choosing a GCP. Check this tutorial: https://www.sarproz.com/publish/data_import_partII.pdf
    If the shift is not global, but local, there can be other possible explanations, like a locally not-accurate DEM.
    Best

    • #2887

      steffan
      Participant

      I will try to select a different GCP than the auto GCP option. What are the best practices for GCP selection at Sentinel 1 resolution? I am aware posts are good for higher res. sensors, but what about ST1? A building?

      • This reply was modified 8 months, 3 weeks ago by  steffan.
  • #2889

    periz
    Keymaster

    First of all, check well the data. Usually Sentinel should have good orbits, not requiring a manual GCP (keeping current orbits option works in most cases). By the way, what DEM did you use?
    About posts, if he post is big enough and isolated (like those in parking lots) it should be firly well visible also in Sentinel data.
    Buildings can be misleading because for a good GCPs the reflection should come from ground level, while buildings may reflect the signal at higher elevation…

  • #2890

    steffan
    Participant

    I am now getting an SRTM error:

    sel_par: neither file D:\SNAP\Beja\Beja_SAR\SLC\2017\RESULTS\DemSynt nor D:\SNAP\Beja\Beja_SAR\SLC\2017\RESULTS\DemSynt.mat found

    files D:\SNAP\Beja\Beja_SAR\SLC\2017\RESULTS\DemSynt and D:\SNAP\Beja\Beja_SAR\SLC\2017\RESULTS\DemSynt.mat not found

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W032.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N04W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N04W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N04W032.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W032.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W033.hgt is all zeros!!

    leggi_srtm: Warning! file D:\SNAP\Beja\Beja_SAR\SLC\2017\..\SRTM\N05W032.hgt is all zeros!!

    An error occurred and was saved in the file lasterror.mat, directory C:\SARPROZ\compiled_2018-02-27 (1)

    If you need assistance, please include the file lasterror.mat

    You may include also file 26_Mar_2018_17_53_35_sarproz.log

    Here a glimpse:

    Error in findtime(): iteration did not converge!

  • #2891

    periz
    Keymaster

    1. if DemSynt is not found it means you have to generate it via “external DEM and synthetic amplitude in SAR coordinates” (this could be the reason for wrong georeferencing)
    2. if an SRTM tile is all zeros there are 2 possible explanations: (a) it’s on the sea (b) the sw tried to download it but there was a failure. If (b) then delete the all zeros hgt files and go through the process again
    3. for future problems (here not needed) include log and lasterror.mat files (notice that u can input your email address and get them by email in case of failure/success of operations)

  • #2901

    steffan
    Participant

    For anyone having the same problem, selecting an appropriate manual GCP seems to fix the problem. Roads and other landscape features are properly aligned now. I used a floodlight from a football field and it seems to do the job just fine.

    The all zero error seems to be due to ocean, as the satellite data does cover some coastal area.

    Thanks for the help.

You must be logged in to reply to this topic.