Coregistration using Manually selected GCP?

Is it possible to improve co-registration results using manually placed GCP with known coordinates?

We had a simular discussion once (Setting ground control points (GCPs)).
The conclusion was that manual points can never achieve the accuracy of cross-correlation.

Depends a bit on your case.
I think it could be useful if manual GCPs could be used for the determination of the initial offset.

I found that cross correlation was not the best suited for my work. So i would like to attempt manual GCP selection and see how the results vary.

Kindly guide me how to do the coregistration based on manual GCP ?

as written in the other topic, there is currently no way to improve the coregistration with manually placed GCPs.

Have you tested DEM-assisted coregistration as well?

Yes I tried that ! but i seem to be having the same results as before.
maybe in my application manual placing of GCP could solve the issue. But i do not understand how to perform it yet .

This is in the “interferometry” sub-forum. I donät think there’s any possibility thant hand-picked GCPs will improve things so that one can do InSAR.

Why don’t you share your use case and we’ll try to help how to solve that exact problem?

Sorry I choose the wrong sub forum. I accidentally choose the wrong one. :confused:

I am currently trying to detect earthquake damage. however my AOI is located that the fault line. So I am able to detect the faults only to one side. After registration when I try compute the coherence chance. one side of the fault line shows the damage however the other side of the fault line has completely low coherence, and thus damage detection is not possible. what should I do in this case ?

If you want to use coherence as a measure for change, the interferometry subforum is suitable. What menhdahl was probably trying to say is that for interferometric approaches, such as yours, manual GCPs are not exact enough, because correct coherence relies on accuracy below the pixel resolution. This can only be achieved by automated registration, such as cross correlation or DEM aassisted registration.

Areas of low coherence as in your example are not a result from bad coregistration, but more likely caused by phase decorrelation over vegetation, for example.
So actually low coherence can be an indicator for change of scattering mechanisms, or even damages.

hmmm that can be possible but i find it odd that only one side of the fault line is showing complete coherence loss ! comparing optically we can see the landslide is only a part of this image.

if the first image was acquired before the landslide and the second was acquired after it is not surprising that the coherence of the landslide area is near zero.

I guess the issue is that the coregistration is done based on one side of the fault line and the movement of the fault decorelated the end result. If you managed to coregister based on the other side of the fault, the coherence-pattern might be reversed, and you could combine the high-cohrence results.

Do you know how much the fault moved?

1 Like

@mengdahl Yes Yes Yes ! Exactly !!! But I am clueless on how to achieve that.

The vertical components of slip ranged from 0.2 to 0.35 m.