Offset tracking

Thanks sir! I really appreciate your information, especially the example. Thanks for your time and consideration!

Dear @hillgrey, Sorry to reply you so late. I haven’t used GAMMA in offset tracking until now. Glad to see you have overcomed this problem.
Best regards.:grin:

Hello sir! Could you please share your experience about GAMMA offset tracking processing?

Hello @Abraun

With offset tracking. What is the minimum unit of displacement that can be identified?

I’m thinking of applying it to the slopes of hydroelectric dams that have displacements.

Thank you.

I’d say this depends on the spatial and temporal resolution of your data. It should provide enough patterns to be recognized in all images but with stable areas used for the coregistration at the same point.

Edit: don’t mind the following, I was wrong with that.
For example, the spatial resolution of Sentinel-1 is originally 3x20 m (range x azimuth, source), which is then resampled to 10x10 m. Accordingly, the observed displacement should be larger than that in some way.

That is not correct, cross-correlation between image templates can easily detect sub-pixel size movements, provided that the tracked features stay similar enough and there is no rotation for example. The window-size can be tweaked to find a compromise between resolution of the output product and SNR.

1 Like

Authors claimed that displacement accuracy of pixel offsets techniques with Sentinel-1 can go up to 50~60 cm in azimuth resolution and 12~16 cm in range resolution. (for highly coherent areas)

That’s quite coarse compared to what you can obtain from Cosmo-Skymed or Radarsat for example. That’s the reason why offset tracking in SNAP is presented for glaciological purposes (plurimetric displacement per year).

1 Like

I wasn’t aware of that, to be honest. Thank you for clarification.

Thanks to everyone for your replies

The displacement accuracy of pixel offsets that I usually see is about 1/20 of the pixel size at best and usually closer to 1/10 of a pixel. As the Sentinel-1 TOPS mode SLC products have pixels that are 14 m in azimuth, I would be surprised if the azimuth pixel offset accuracy could be less than 0.7 meters for most situations.

2 Likes

Hello, ABraun,
The data in the figure below is dual polarization (VV + VH). Do I need to choose one of them when doing offset tracking?
image

VV should work better in most cases due to a better SNR (signal to noise ratio).

1 Like

@mengdahl Which operator in SNAP should I use to separate them? Thank you.

During the radiometric calibration, you select only the VV or HH band

I have a question. In addition to SNAP, GAMMA can calculate the pixel offset, is there any other open source software can achieve this method?

The Open Source ISCE2 software includes functions to calculate pixel offsets from SAR images. It is on GitHub:


The ISCE2 software is command-line only and runs on Linux and Unix-based operating systems, including MacOS. Microsoft Windows is not supported.

3 Likes

@EJFielding Do you have any information on setting up ISCE2 on virtual machine? It was done for snaphu in the past years.

There are positive and negative values ​​in the offset tracking velocity results, so what do they mean?

@mengdahl
(1) What algorithm is used for offset tracking in SNAP? Amplitude strength tracking method or phase coherence tracking method?
(2) Is there a corresponding reference, which may be needed in my thesis?

Amplitude. The module only reads detected images.