sarproz-group

use of Graph Analysis and Refinement

This topic contains 6 replies, has 2 voices, and was last updated by  bridgetwang 4 months ago.

  • Author
    Posts
  • #4049

    bridgetwang
    Participant

    Hi,

    When I use the Graph Analysis module as well as read the manual, I come up with the following questions:

    1. What the “Skip init Mask” means? When should it be checked?

    2. As the NOTE 4 in manual says:”NOTE 4. in this module you can optionally save a time series object. However, for this purpose the atmosphere should have already been estimated in the APS module”.

    I can’t quite understand what NOTE 4 means:

    (1) What the time series generated in this module is. Does it means that “Sparse Points Processing” step can be replaced by this step?

    (2) Also, NOTE 4 emphasis that in this case APS estimation must be runed. But shouldn’t APS estimation always be done before Graph Analysis? Is it possible a way to skip APS?

    Thank you in advance for answering !

    Best
    Zoe

  • #4055

    periz
    Keymaster

    1. What the “Skip init Mask” means? When should it be checked?

    it means selecting all pixels, not loading the initial mask
    https://sarproz.com/manual/set_mask.html
    https://sarproz.com/manual/carica_maschera.html

    What the time series generated in this module is.

    a time series object is a file that can be saved either in the APS module or in the MISP module (as well as in the “graph analysis and refinement” -GAR-) and that can be loaded in the “time series” module
    https://sarproz.com/manual/ts_module.html

    Does it means that “Sparse Points Processing” step can be replaced by this step?

    yes, this is an alternative. You can use the GAR instead of the MISP. As an example, there are cases where the MISP cannot be applied because of too strong movement. there, you can use the GAR to densify points. We will publish a tutorial on such cases in the near future.

    Also, NOTE 4 emphasis that in this case APS estimation must be runed. But shouldn’t APS estimation always be done before Graph Analysis? Is it possible a way to skip APS?

    you cannot skip the APS estimation and removal if you want to produce time series. But you have 2 options for using the GAR:
    A. you use the GAR only to densify points. So, firstly you run the APS module, you estimate the APS, then you densify points with GAR
    B. you run the APS module. you are not happy with the graph inversion (the APS is bad). Then you run the GAR and you improve the graph. then you go back in the APS module, you load the improved graph, and you estimate the APS delay again.

    SUGGESTION: if you want we can work together at organizing a course in your region. We can send someone there to give a 5 days intensive training, with theoretical and practical classes. This would help you a lot understanding many things. Starting from the fall we’ll also be able to organize 1 week training courses in Italy -and for that no need to gather people in your region-.

    best

    • #4123

      bridgetwang
      Participant

      Dear Prof. Daniel,
      After I read your answers, I tried some experiments with GAR to comprehend better in the past days. I got some more questions here during the experiments:

      1. Does GAR always behave better than MISP? And do you recommend to use GAR in most cases?

      2.You mentioned GAR could detect strong movements which exceeds wavelength, I want to know what the limit amount of moving GAR could detect? why could it detect movements larger than wavelength?

      3. Are the “local reference points” in GAR still PS points, or are they other points like DS points?

      4. With respect to your suggestions, I’d like to ask has date of the course in Italy been set yet ?

      Best regards,
      Zoe

  • #4129

    periz
    Keymaster

    1. Does GAR always behave better than MISP? And do you recommend to use GAR in most cases?

    the standard way APS+MISP is widely used and very robust. when APS+MISP is successful, you don’t need anything else. However, there are cases where it fails. In those cases, GAR can help

    2.You mentioned GAR could detect strong movements which exceeds wavelength, I want to know what the limit amount of moving GAR could detect? why could it detect movements larger than wavelength?

    If A and B are faraway and move a lot with respect to each other, you won’t be able to process them correctly. However, if you have C and D in between, maybe you can analyze the path A – C – D – B successfully. In the MISP, assuming your reference is A, you can only analyze A – C, A – D, A – B. In the GAR, you can analyze A – C – D – B and, if needed, even add more pixels that will be connected to the closest stable point. For the rest, the usual limitations/constraints related to the wavelength still hold.

    3. Are the “local reference points” in GAR still PS points, or are they other points like DS points?

    they are selected among the best points previously analyzed

    4. With respect to your suggestions, I’d like to ask has date of the course in Italy been set yet ?

    In Italy, starting from the fall, we will be able to organize training courses on demand, there won’t be just a single course. However, we’ll need to gather participants. At the moment we don’t have a date yet to collect applications. We are also talking about a training course in China though. Please send an email with your wishes and preferences.

    best

  • #4133

    bridgetwang
    Participant

    Thank you for your kind reply.

    With respect to question 3, the “previous points” you pointed to are points selected by e.g. Amplitude dispersion index in APS?

    Also, I have the confusion that, since QPS make use of DS points, shall I comprehend like this: The points selected with star graph are PS points, and points selected by other kind of graphs(MST,Full graph…) are PS+DS points?

  • #4136

    periz
    Keymaster

    With respect to question 3, the “previous points” you pointed to are points selected by e.g. Amplitude dispersion index in APS?

    the GAR works with a graph previously processed by the APS. so “previous points” are the points that you import in GAR from the APS module

    The points selected with star graph are PS points, and points selected by other kind of graphs(MST,Full graph…) are PS+DS points?

    you do not select points via an images graph. you select points based on some parameters and/or conditions. what is important is weather your point is coherent or not in one images graph or in an other. this makes them ps or ds. if a point is coherent in a star graph configuration, no need to change the graph

    best

  • #4138

    bridgetwang
    Participant

    Thank you very much !

You must be logged in to reply to this topic.