Astro Pixel Processor

Mosaic Integration Error:java.lang.NullPointerException

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #15226

    Rosen
    Participant
    posts: 16

    Mabula – I’m having problems with integrating the mosaic I described in the 1.038 thread – it’s a 14 panel mosaic from the Canon 6D and the same 50mm lens that I used for the North circumpolar data. This one will end up having the vertically oriented panels side by side, so they should create a mosaic of roughly 210 x 40 degrees when I’m done.

    I’ve tried a number of different ways of integrating the data, both full stacked frames as well as from stacks that were cropped (to eliminate some of the corner coma of the 50mm lens at f/2.8). With every method I try, I end up with an error when I do Integration and it always happens at 88%. Here is a screen shot:

    Here’s the basic process I’ve used:

    Each panel is a stack of 20 x 10 second captures on a fixed tripod. So, they each have some trailing of a handful of pixels. However, I don’t think that’s creating a problem for APP, as I was able to mosaic 3 of the panels. I had some registration problems, but it didn’t bomb out.

    Each of the files was calibrated using Images Plus (old habits die hard).

    In APP, I loaded the 14 panels

    In 3) Analyse Stars -I used all the defaults.

    In 4) Register – I set the reference panel to the middle panel of the mosaic. Pattern Recognition descriptors were all defaults (pentagons – I also tried quadrilaterals, scale start 1, scale stop 5, no flip x/y descriptors). I choose “use dynamic distortion correction” and “same camera and optics”. I had it create a distortion model, # of overlapping frames 3 (not sure what this does). Registration mode is Mosaic, registration model is calibrated projective. I then clicked “start registration”. After that, I did not save the registered files.

    In 5) Normalize – mode is regular, projection Equirectangular projection, method add-scale, scale BMWV, neutralize background is checked. I then clicked “normalize lights”.

    In 6) Integrate, weights – equal, integrate – average, mode – reference, I did not choose a new reference frame. For LNC degree, I chose 2nd degree LNC, iterations of 3. MBB was not enabled, outlier filter is no rejection, pixel interpolation filter is lanczos-3, no under/overshoot was not checked. Integrate mode is no drizzle, scale 0.3, projection is equirectangular projection. I clicked the Stack button, it takes awhile (mostly while on LNC iter 1 performing LNC calculation while it says it’s completed 30%). It appears to complete LNC, quickly goes from 30% to 88% and then comes up with the error message.

    I’ve tried varying a handful of options, but still end up with the same error. I’d appreciate it if you could give me any more ideas to try or let me know if you need any more information.

    • 1 person likes this.
    #15227

    Rosen
    Participant
    posts: 16

    A followup to this message.  I think the problem I’ve been having may be related to the # of overlapping frames.  I’ve been able to get past the message by choosing the same number of overlapping frames as the total number of frames to be integrated.  So far, I’ve been able to integrate 3, 4, and 5 panels.  I tried doing 9, but it failed with a different error (I didn’t take note).  And, I don’t have an option to choose more than 9 panels, anyway.

    Now, I say I’ve been able to integrate 3, 4, and 5 panels, although that means that it’s completed without any errors.  However, it’s not really mosaicing these panels (and the registration isn’t good, either).  As I add panels, the mosaic should grow vertically (I’ll end up rotating it 90 degrees).  Each of the stacked frames as an aspect ratio of roughly 3:2.  So, when I have 2 panels it should be 3:4 (it’s less because of overlap), 5 panels should have an aspect ratio of 3:10 – with some overlap, maybe 3:7 or so.  However, virtually all of the integrated mosaics still have aspect ratios of 3:2.   So, I think I’m still doing something wrong.

    Maybe it’s not handling the slightly elongated stars from the 10 second untracked 50mm subs?

    #15228

    Rosen
    Participant
    posts: 16

    Ok, I’m sorry to keep wasting bandwidth on half solving my own problems, but it’s been a learning experience at least!  I just found that my problem with the mosaic not expanding as I add frames is that I had chosen a mode of “reference” in 6) Integrate.  Changing that to Full solved this problem as well as the majority of the registration problems.

    I’ve successfully integrated 4 panels.  I’ll try to do more and report back later.

    Sorry again for all the unnecessary comments.

    • 1 person likes this.
    #15234

    Haverkamp
    Participant
    posts: 648

    Hi Scott (@srosenfraz),

    The first bug that you describe is a new one for me. I know that at 88% the stack should be finished. Then from 88% to 100% APP performs analysis of the stack and saves it to your harddrive.

    So it seems that the stack possibly isn’t there, possibly to a problem in registration.

    If you could share these panels, so I can duplicate the error over here, I can probably fix this quickly. (You can use our dropbox folder.)

    Regarding the mosaic settings:

    You can select scales now, use scales of 5-10 for the mosaic integration. Try to have at least 2000 stars detected per frame with the settings in 3)

    deselect “same camera and optics”

    “dynamic distortion correction  “on

    and calibrated projective mode and mosaic mode.

    The overlap number refers to how many times a single panel in your mosaic would overlap with other frames. For a big mosaic, probably a value of 6-8 would be needed. But it will take some time to calculate. I previously had this fixed at three, so a higher value would be more robust if the mosaic panels overlap wit more than 3v panels each.

    Indeed, use the full composition mode to include all pixels of all frames in the integration.

    Regarding, the calibrated projective mosaic algorithm, that one still needs a big! upgrade. I hope to do this within the next week or so. Probably, it’s best to try first, for now, parts of the mosaic with the “uncalibrated” projective model. This algorithm already received a big upgrade last week.

    (my todo list keeps growing 😉 )

    #15247

    Rosen
    Participant
    posts: 16

    Hi Mabula –

    I wouldn’t worry too much about the error I was receiving. I eventually found that the problem was coming from trying to integrate the mosaic with the Integrate mode of Reference. Once I changed that to Full, I stopped receiving those errors and the combined panels were no longer ending up the size of a single panel. So, this one was entirely operator error.

    One suggestion – I imagine there isn’t a scenario where someone would choose Mosaic in 4) Register and then want to use Reference in 6) Integrate. Perhaps it would be helpful to have a sanity check in Integrate that pops up a message if someone has chosen Mosaic and then Reference – something like “are you sure you want to use Reference? Mosaics usually should be integrated with “Full”.

    I did find that APP could complete the integration if I chose overlap of 3 panels (in this mosaic all of the panels touch a panel on either side, except the end panels only overlap with 1 other panel).

    As far as the 88%, what I was seeing with the error was that APP would try to do the 1st LNC correction. At the end of that initial pause, it would NOT go on to the normal subsequent LNC steps. Rather, it would advance from 30% to 88% and then issue the error. Again, this is a user triggered error, but I thought you may want to know the sequence where it was occurring.

    I’m still having some problems with this Mosaic in having the panels properly register and then have the projection look correct. These 14 panels are portrait oriented side by side along the horizon line for about 210 degrees (give or take). I’ll be combining them with a similar mosaic of the landscape below this to make a panorama of my surrounding mountains with the night sky.

    Using Equirectangular or Rectilinear projections ended up creating significant warping of the panels on the ends of the mosaic. I was able to use Photoshop’s Photomerge with a Cylindrical projection, and that seemed to align the panels as I expected. However, I’d certainly be interested in finding out how I might do something similar with APP.

    I know you’re very busy, and this is not a priority in any way. But, I did upload the 14 panels (I reduced them to 50% size, but they’re otherwise unaffected. I also uploaded a jpeg of the Cylindrical projection so you can see what it should look like.

    • 1 person likes this.
    #15264

    Haverkamp
    Participant
    posts: 648

    Hi Scott,

    Thank you for uploading the panels, I’ll have a look tomorrow and will give you feedback 😉

    Mabula

    #15761

    Haverkamp
    Participant
    posts: 648

    Hi Scott,

    Apologies for my late response ;-(

    “Perhaps it would be helpful to have a sanity check in Integrate that pops up a message if someone has chosen Mosaic and then Reference – something like “are you sure you want to use Reference? Mosaics usually should be integrated with “Full”.  ”

    This is now fixed in 1.043 which I’ll release tomorrow or monday.

    I’ll still perform testing using the data that you sent me.

    I think for such mosaics, I need to do more upgrading in the mosaic algo’s and need to add more projection methods.

    Cheers,

    Mabula

Viewing 7 posts - 1 through 7 (of 7 total)

You need to log in or to reply to this topic.