that is fine then.
I really hope so. I have just an 8GB RAM probably that might be slowing the process.
Hello @ABraun! Thank you for all the help. I have been able to process the pair I was working on. I tried out the Displacement just the way you said. Now there are some values in the colour manipulation section which are negative ones. Those lie in the topmost section of the colour chart when it is exported in the Google earth interface. Does that mean there is least displacement here? Other than that the image has turned out to be more or less according to what I expected.
beware of outliers in your displacement map. They mostly result from atmospheric effects or unwrapping errors.
More important is the overall pattern of your displacement map. Depending on the nature of your displacement it concentrates on certain areas.
Alright. But how to understand whether I have outliers in the displacement map? Well, I tried to check the map out in google earth, it seemed okay. Any suggestions on how to identify these errors?
Well, you probably know your study area and can estimate where displacement is reliable and where not.
If the fringes show nice patterns, there is nothing wrong. But if you see a trend in your data superimposing the fringes or have some noisy smaller areas, these are probably not of use.
Always look at coherence and mask out low coherence areas to exclude unreliable pixels. Have a look at this example: Subsidence map in 3d view
Thank you, I understand what you suggest @ABraun. I’ll try out the coherence as you have suggested and mask out the unwanted pixels. Once I complete I shall inform you promptly. But just a question, if I have to clip out some fringes, according to my requirement, should I carry out the coherence math and then clip it out?
the coherence mask is just a visual thing. You can crop out some parts of your image and then apply it again. This, however, requires that the coherence band is present in the product.
Okay. One more query, during the filtering phase I used the coherence mask. So do I have to do it again? Using band maths?
yes, it doesn’t erase the pixesl permanently. It just excludes them from the view, but it helps to interpret the interferogram.
I understand. But I am not being able to do the coherence mask. After going through with the equations, when the mask is done, very few places are being masked. Nothing major. So is there something wrong?
what is your threshold?
How does the initial interferogram look like?
I used 0.3 as my threshold. The interferogram is not distinguishable at first with a particular pattern. But a pattern can be seen after unwrapping the image.
hmm, October 2007 to July 2009 is quite long.
Can you show an RGB image of the two intensitiy layers coregistered stack?
In the meanwhile I thank you for your useful advice.
your image doesn’t look like the coregistration was sucessful. Can you make one just with HH of both dates?
This would explain why there are no fringes in your interferogram.
Okay. I’ll give it a try. But I found a pair more closer to each other according to the time scale. Will that do? it is from July to October 2007.
The coregistration process is almost automatic, once the correct options are chosen. Of course I chose cubic convolution 6 points as the interpolation method. Should I choose something else then?
Yes. First image is red, second image is green. No blue. It is important to check that both images are perfectly aligned. Based on your previous images I’m not sure about that.
I understand. Well, there is a portion to the left which is missing in one imagery. Mostly towards the extreme left. Other than that this is the coregistered stack, in RGB with only two HH imageries of the two dates without any selected in the blue section. Please tell me how to go about the whole thing correctly. I have been trying every possible way to go about this job of mine
that doesn’t look correct.
Go back to the co-registering step and increase the number of GCPs to 3000 or 4000, activate the coarse coregistration and reduce the GCP tolerance to 0.2.
If this doesn’t bring a better result, try the DEM based coregistration instead.