Only 11 of 48 pics recognised -> rookie question

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • Michael
    3Dflower
    • Mar 2021
    • 3

    Only 11 of 48 pics recognised -> rookie question

    Hi all,
    I am completely new to Photogrammetry, so please bear with me if this is a dumb question:
    I have started the program, all setting within 3DF were left standard and I just clicked through without changing the default settings. Now I imported my first project (48 pictures) which I photographed using a 3d scanner (Openscan Classic from Openscan.eu site).
    This being said, the camera is steady and the scanned object moves around in front of the camera, so this is where my questions begin.
    The result is like the screenshot below. Besides that only 11 of my 48 pictures are recognised, I do not know why the camery positions are lined up as shown below... how can I avoid that (camera position is not correct as I would assume from below output) and tell that all 48 pictures should be processed...?
    TIA!
    Michael
    ?
  • cam3d
    3Dflover
    • Sep 2017
    • 682

    #2
    Hi Michael -

    Without seeing the source images it's hard to say what exactly is going wrong, but my suspicion is that you haven't masked out the background of your images, and that 3DF Zephyr is finding strong feature matches in the background of your photographs. These feature matches override your subject feature matches and pile all your images on top of eachother in the viewport.

    Masking is a great way to isolate relevant data. Here is an article you might find helpful: https://www.3dflow.net/technology/do...for-turntable/

    Comment

    • Michael
      3Dflower
      • Mar 2021
      • 3

      #3
      Hi cam3d,
      thanks for pointing the mask feature out. This helped a lot!

      I ran over the tutorials and have now a slightly better result - but far away from being perfect :-(
      I will try to explain what I did with a small sample:
      first photo shows the object whith turns around - not the real object of course :-)
      second shows how I did the masquerade (I did it for each photo separately to be more accurate).
      Calibration was done automatically (GT-i9300 which is a Samsung S3 mobile)
      Everything else was left at default.
      Third photo shows intermediate result just before I hit run.
      fourth photo shows that only 12 of 48 photos could be processed - which is far too less.
      last photo shows the outcome, which is not the complete object. May be the pink surface being to glossy?
      Any hints on that?
      TIA!
      Michael
      Last edited by Michael; 2021-03-22, 11:14 PM.

      Comment

      • cam3d
        3Dflover
        • Sep 2017
        • 682

        #4
        Michael I'm glad the masking helped!

        Something to note is that your subject is feature poor (it's a smooth plastic surface with very few distinct features for Zephyr to work with) - If you were to add some features using a powder (talcum powder can be good!) then you should see that your 3D reconstruction improves a lot!

        Comment

        • nthnlwilbourn
          Blossoming 3Dflower
          • Mar 2021
          • 1

          #5
          Originally posted by cam3d
          Hi Michael -

          Without seeing the source images it's hard to say what exactly is going wrong, but my suspicion is that you haven't masked out the background of your images, and that 3DF Zephyr is finding strong feature matches in the background of your photographs. These feature matches override your subject feature matches and pile all your images on top of eachother in the viewport.

          Masking is a great way to isolate relevant data. Here is an article you might find helpful: https://www.3dflow.net/technology/do...for-turntable/employee monitoring software
          thanks a lot, the article is really helpful. I have seen similar problems with image recognition many times already, I hope to fix it in the near future

          Comment

          Working...