sarproz-group

Graph Refinement Error

This topic contains 4 replies, has 2 voices, and was last updated by  periz 4 months ago.

  • Author
    Posts
  • #2695

    janbu
    Participant

    Hi,

    I have a few general questions on the Graph Refinement and a specific error that I got loading the connections back to the APS module.

    – Should we keep points parameter and value exactly the same as in the previous APS (that’s what I have done?
    – The Graph Refinement Window also gives the possibility to change the processing parameters. Of course, I kept the same as APS. However, In which case will be those modified within the Graph refinement module?
    – Once I upload the SPRMA nack to the APS I get an error (“index exceeds matrix dimensions”) and instead of connections, only points are shown (image attached).
    – once we move to the MISP, can we densify again, (if it is the case), by decreasing the parameter threshold (eg: ampstab ind)?

    Thank you for your support

    Attachments:
    You must be logged in to view attached files.
  • #2698

    periz
    Keymaster

    – Should we keep points parameter and value exactly the same as in the previous APS (that’s what I have done?

    * are you talking about the “points densification” section? no. you can choose any parameter/value. in fact, the idea here is that you can add as many points as you want. You have already run the APS module. So, you have processed connections. Thus, you know which points are good. Adding points here means keeping good previously processed points as local references, and add the new ones as connections to those good references.

    – The Graph Refinement Window also gives the possibility to change the processing parameters. Of course, I kept the same as APS. However, In which case will be those modified within the Graph refinement module?

    * this is something we did not explore yet. But the concept could be explained by the following example. Let’s say we are in a urban area and we want firstly (in the APS module) to connect points on the ground (without considering e.g. the thermal expansion). And later (Graph refinement) we add also points on buildings. In this case, in the “graph refinement”, we may want to estimate also the thermal expansion. In principle, you could do that (choosing different processing parameters). In practice, we did not try it yet and we may need to make a few little changes to make it doable.

    – Once I upload the SPRMA nack to the APS I get an error (“index exceeds matrix dimensions”) and instead of connections, only points are shown (image attached).

    * yes sorry, this was a bug that we corrected yesterday. you have to download the last version and run again the Graph refinement. The SPMRAutosave is missing a piece.

    – once we move to the MISP, can we densify again, (if it is the case), by decreasing the parameter threshold (eg: ampstab ind)?

    * you can do what you want. but since you can already densify points in the “graph refinement”, I would not use the MISP to densify again. Please be aware of the conceptual difference: in the graph refinement you integrate a spatial graph. In the MISP you re-run the analysis using 1 single spatial reference point.

  • #2699

    janbu
    Participant

    Thank you for the prompt answer. That was very helpful.
    – Regarding the first question/answer, is that valid also if after the Graph Ref (GRA) we recompute the atmosphere in the APS? Is that recommended (or when it is recommended)?

    – I noticed that when we upload the autoconnex to the GRA, by default a connection coher threshold is populated. In may case was 0.63. Does it make sense to increase it (in order to create more robust clusters? In my case, I set 0.88 (see images attached below. Should I keep lower (as default)?

    In the GRA I have densified the points by lowering the threshold to 0.7 and skipping the initial mask. After the GRA, I have recomputed the APS. The parameters seemed OK.

    Unfortunately, in the MISP, the deformation (and vel) parameters have a strong/unrealistic variation, I presume along junction of the 2 frames (sentinel). In a couple of previous standard precessing (no GRA) I didn’t have that effect.
    What is it happening here? Are my steps/reasonings right?

    Thank you in advance

    Attachments:
    You must be logged in to view attached files.
  • #2704

    janbu
    Participant

    Sorry…file named GRA4 is actually in the APS module (when reading the SPMRAutosave).

    Here are attached the following the images. MISP7 shows the “error” in the deformation pattern.

    Cheers

    Attachments:
    You must be logged in to view attached files.
  • #2708

    periz
    Keymaster

    – Regarding the first question/answer, is that valid also if after the Graph Ref (GRA) we recompute the atmosphere in the APS? Is that recommended (or when it is recommended)?

    * if after the graph refinement you import the SPMRautosave in the APS module, you do not need to estimate parameters anymore. you have already estimated them in the graph refinement module (where by the way you could even export them). If you import the SPMRautosave in the APS module, you do it only for estimating the APS.

    – I noticed that when we upload the autoconnex to the GRA, by default a connection coher threshold is populated. In may case was 0.63. Does it make sense to increase it (in order to create more robust clusters? In my case, I set 0.88 (see images attached below. Should I keep lower (as default)?

    * the sw always gives a default option for those who have no means to find a better one. You surely can increase the default coherence threshold given by the software in order to make the inversion more precise. However, should the inversion of the spatial graph fail, you would need to consider lowering the threshold back towards the suggested value.

    – In the GRA I have densified the points by lowering the threshold to 0.7 and skipping the initial mask. After the GRA, I have recomputed the APS. The parameters seemed OK.

    * unfortunately a vertical line is visible. this is clearly a phase jump problem related to sentinel. you have to check the log (and/or send it to yuxiao).

    – Unfortunately, in the MISP, the deformation (and vel) parameters have a strong/unrealistic variation, I presume along junction of the 2 frames (sentinel). In a couple of previous standard precessing (no GRA) I didn’t have that effect.
    What is it happening here? Are my steps/reasonings right?

    * see the previous answer.

You must be logged in to reply to this topic.