[cnuhyl]

Hi,
When I make the sparse points processing ,there has a warning:Matrix is singular to working precision.Does it is a error? And it whether or not has some influences on the result？
Thank you.

Attachments:

You must be logged in to view attached files.

[periz]

I looks like the following:
– you are processing a multi-master images graph (it would be nice to see it)
– you are using the coherence as a weight in the inversion (is this the right choice? you need to look carefully at all your interferograms)
That message says that there are points for which some low spatial coherence values break the images graph and this makes the time-inversion unstable.
So, fur sure for those points the results will be unreliable (but, as you see, the sw is not stopping: the message is a warning, not an error).
First question: is this happening for a few points or for many of them?
Then: did you look at the result? You need the right criticism to judge whether this is strongly affecting the final result.
Consider anyway that it is something affecting single-time series and not the result as a whole (in sparse points processing points are analyzed one by one independently from each other)
At this moment Sarproz uses the SVD for inverting the time series only if the graph is disconnected since the beginning. But we could add a manual option for such cases. We will consider it.

[cnuhyl]

Thank you very much.
when I processing a multi-master images graph ，what is the different about using coher or Amps as a weight? which is better?
Should the MST graph all connected? And how many connections better?

[periz]

There are no better or worse things. They are simply different and they apply to different analysis/objectives.
Using the Amplitude (Models!) is useful e.g. in the case of temporary targets. You don’t use it in multi-master approaches.

The MST is by definition connected. However, even if theoretically interesting, the MST is practically not very robust. Better to have redundant connections. You can create a redundant graph e.g. by placing a threshold on the coherence.

[Author]

Posts