Offset tracking: when does the GCP have no valid offset?

Dear experts, is it right to believe that there are three exact cases when the GCP have no valid offset and a hole arises? (1) an estimated velocity exceeds the maximum allowed velocity, (2) cross-correlation is less than the threshold, (3) cross-correlation is high enough but an estimated velocity equals zero.

quite theoretical qustions - could you please provide examples? You can upload screenshots in here.

(1) and (2) are privileged. You can verify (1) by setting a max velocity value super high and compare with your “normal” max value. Then you can see where scenario (1) occured.

Scenario (3) is very unlikely to occur. With the oversampling factor, post-filtering and so on, you will always find a non-zero value. Personal experience confirms this. Note that the color used by SNAP makes very low values almost transparent, it does not mean that there is no value.

In the codes, the value of the correlation is computed (in the getOffset function, lines ~800-900). It would be super useful if the module would export the correlation as an image (the same way that coherence is an image in interferometry). I tried to modify a bit the code in order to do so but some Java limitations don’t let me do it as easily as what I wanted. I don’t know if exporting the correlation value can be requested for near future. Maybe experts @lveci can give us a thought?

1 Like

Yes, it would be great to have correlation image.
In my case I unchecked ‘fill holes’, made offset tracking and SNAP showed me some ‘empty areas’ with zero velocity. My assumption was that there were no movement in these areas. Today I have repeated calculation with max velocity 1000 m/d and correlation threshold 0,0000001. The holes have been disappeared. Scenario (3) is really unlikely.