Suitable parameters for offset tracking

My working topic is to determine the sand bar migration pattern around the Padma Bridge, Bangladesh.From sub pixel correlation it is found that the displacement is about 150 m in 17 years. From offset tracking i am not able to get the expected result because of the lack of knowledge on selecting proper offset tracking parameters.
The parameters include :Grid Azimuth spacing ,Range spacing, window width, window height, cross correlation threshold, average box size, max velocity(m/day),radius for hole filling
Plz help me out by providing the suitable parameters. Thank u so much in advance.

if there were ideal parameters for all kinds of sensors, resolutions and applications, there was no need for manually entering them :wink:
Do you really think we can simply provide you a bunch of numbers which exactly fit to your study? You didn’t even mention what data you are using which steps you undertook by now ect. Also screenshots help us to understand your problem. But we cannot know unless you tell us a bit more.

I’m using a lot offset based techniques to integrate surface velocity fields into geophysical models.

Offset tracking is often used to study glaciers because of their fast movements (>100 m per year). In your case, i’m not even sure that your displacements are important enough (<10m per year) to detect movements.

I recall that offset tracking works by finding an offset in pixel which maximizes local correlation. It is thus applicable to detect at least pixel-side displacements (especially in SNAP). For your application, it means that you need very large temporal baseline (> 1 year) to have a displacement large enough, meaning also that you can probably dramatically increase the temporal decorrelation.

Please, give more information about your images, your results, and so on :slight_smile: In addition, please refer to the SNAP tutorial related to offset tracking :

1 Like

Thank u so very much…yeah i know its usually prefered for fast movements.But now um trying in river side…Anyway, um using S1A_IW_GRDH_1SSV for 2014 & 2017 (SNAP). And um also following the given tutorial.But there is no explanation for determining the parameters. By default parameters are used here and the the result is attached bellow…
pa

I think 2014 to 2017 is too large for SNAP to be able to detect the offset. The surfaces probably changed very much during that time.
The module is ideally designed to find small shifts in patterns of surfaces. So, as many Sentinel-1 data are available in the archives, I would rather suggest that you use image pairs at a monthly base to test if a small offset within the reasonable range (and direction) can be observed. Then you can gradually increase the time span between your image pair to see if it also works.

1 Like

thank u for ur suggestion…i will try…

In your parameters, you let the Max Velocity value as 5m/day which is faaaaaar beyond what you told me (150m/17years). This parameter is usefull to discard outliers.

I don’t know if your results are correct, but SNAP found some. I can’t see the values. Are the amplitude of the velocity coherent with previous studies?

As @ABraun said, begin with small temporal baseline to minimize temporal decorrelation. Then, depending on the results, you can increase. Think about your concerns as an optimization problem