S1 TOPS Coregistration "Cannot Create Data Buffer"

[DanielaPL]

Hello S1 people:

I am trying to coregister two imags following the tutorial of ESA and while it runs it says that it cannot create databuffer and it stops. Why is this happening? what should I do to solve it?

I have a Windows 10, 8GB memory. and the version 4.0.0 of SNAP.

I have done somethinf under tools-options where I checked an option about saving memory, as I read it in another post. However, the problem persist.
Thank you.

[hubert]

Hi DanielaPL,

Are you process whole scene (3 IW)? Unfortunately SNAP 4.0 (as also 3.0) have some problems with RAM memory optimization. There is no way to process whole scene of S1 images unless you have 32GB RAM memory. It is strange that older version (Sentinel-1 Toolbox v.1.1.1) haven’t this problem. So there is some idea to use Sentinel-1 Toolbox v.1.1.1 instead of SNAP 4.0.

[DanielaPL]

@hubert:

Thank you very much. I was able to run it on my computer, but it took ages.
I was wondering if you could help me with some doubts I have regarding co-registration of S1 IW images:

1: What are the best parameters to use regarding the co-registration method and the re-sampling method.
2: After co-registration how can I measure the accuracy of the process? Is there a file I can read where I get a RMS or something.
3. As far as I understand. A co-registration process creates a stack with both images (master and slave), can I have the coregistered and resampled salve image on its own and not part of the stack? This is to do coherence analysis between different pair of images. lets say a slave+slave. The coherence estimation tools seems to only work with stack images. Is there a way to stack pair of images together?

I am following the ESA tutorial for interferometry, but I need to do more coherence analysis that Interferometric Analysis. Do you know how I can use other software, for example R, to make some statistical analysis of the data?

Sorry if my post is too long. I have some many question, because this is the fist time I deal with RADAR data.

Kind Regards,
Dany

[falahfakhri]

Hi,

Please take a look at this

Terrain-Correction: GC overhead limit exceeded

You could try to break it down into two graphs. After mulitlook put the GRD-post processing which just renames the product and changes its product type and write it, then terrain correct as a separate step which is probably the heaviest. Also, update georeference is only intended to update the geocoding in the product similar to terrain correction but keeping the data in SAR geometry and instead creating a pixel geocoding of where the lats/lons would be if the product was terrain corrected. Thi…

Answering your both questions in one the good parameters and the
scrutinize of the result could be checked via the range and azimuth RMSE

[DanielaPL]

Thank you. I was able to run the co-registration process, but can you be more specific about the last 2 lines you wrote?

“Answering your both questions in one the good parameters and the
scrutinize of the result could be checked via the range and azimuth RMSE”

I am not sure I understand.
Thank you for your time

[falahfakhri]

Hi,

Would you please to take a look at the following link at the paragraph of Corregisteration

FAKHRI F. (2013) Long and short term... (PDF Download Available)

[DanielaPL]

@falahfakhri

Thank you for the interesting document, however. it still does not answer my question on which parameters should I use with the SNAP software and where does the SNAP software stores the accuracy results of the co-registration process. This document uses GAMMA and not SNAP.

Thank you.

[falahfakhri]

Hi Daniel,

I think Mr. Luice has answered you via the following link,

DEM Resampling

The Bisinc resampling is probably best but slower. For most cases a simple and fast bilinear should be good enough. For cross-correlation based coregistration, you can view and export the RMS values through the InSAR Stack Toolview found in View -> Tool Window -> Radar -> InSAR Stack For TOPS coregistration, we don’t have RMS values but you could look at other statistics such as for the ESD.