[paulinanino]

Dear all,

I have some doubts about a project I am working on. I am using 25 interferometric images to measure deformations and in the “APS processing” stage I would like to use the filter “ampl. stab. index” > 0.8 since I have few photos, however, I get very few points especially in the area I am interested in (red rectangle in the attached image). The only way to get points in that area is to use ampl. stab. index > 0.6 but I am worried of getting bad results. I tried using another (filter “refl. map” > 0.8) and it’s the same.

On the other hand, in the “non linear weighting” part I am setting the m=0.8 and M= 0.97 is it ok? how can I be sure that I choose good values?

Thanks

Attachments:

You must be logged in to view attached files.

[antig]

APS is long wave length signal. So 0.7 seems fine spatial distribution for aps. In sparse point protssesing u may use lower ASI 0.6 for exaple

• This reply was modified 9 months, 1 week ago by antig.

[paulinanino]

Thanks, your help was very useful!

[periz]

Hi Paulina,
there are multiple things to consider in your case. An exhaustive answer to your question would require to discuss about various topics. You might want to take a Sarproz course to get more experience (in case, contact courses@sarproz.com). For the moment let me tell you that you should consider the APS estimation as an independent process with respect to the estimation of the parameters of interest. It means, firstly focus on the APS estimation with the points you have, after that, you could try to make a denser estimate in the MISP module (or in the Graph Analysis and Refinement).
best

[paulinanino]

Okay! I understand much better.
And for the non linear weighning part, what would you recommend? I understand that the lower the number of images, the more coherence is required to trust the data, so m=0.8 would be a good value when it comes to few images, is that right?

Thanks!

[periz]

you are right that a low number of images implies low robustness in the estimation of the coherence and a bias towards high values. That said, numbers are relative and dependent also on other factors. so, it is difficult to tell you what number to use without looking at your dataset and at your results in detail. play with the numbers and observe the consequences.
best

[Author]

Posts