Aligning Cameras in RC vs Photoscan

Answered
Follow
Avatar
julesmalcomson
Hi,

I have a load of head scans to process shot on a multi-camera rig (40cams @18mp).
I have found (as I have found many times before) that RC struggles to align all of the cameras, while tests in photoscan are all 100 percent successful.
I have played around with the alignment settings before but I still don't entirely understand the controls.
Alignment time is not really an issue for me, is there a way to increase the settings to get more reliable alignment?
Or can someone explain the relevant parameters to achieve better success aligning images?
As it was a multi-camera array the alignment the cameras are all fixed and were shot simultaneously, so it seems the issues were down to variations in reflections from one image to another.
Thanks
0

Comments

6 comments

Sort by Date Votes
  • Avatar
    Wishgranter
    Hi julesmalcomson

    I highly recommend you to create one perfect alignment (on shooting start) with something like mock-up head with plenty of non-uniform patterns like newspaper pattern. Then set scale orientation etc.
    And use XMP workflow for transferring the alignment to other data, so you do not need to realign and will not get into trouble.
    Have you seen the XMP workflow or do you need my assistance? Can you prepare a good alignment dataset, so that I can instruct you further?
    0
    Comment actions Permalink
  • Avatar
    julesmalcomson
    Thanks for the response,
    I have never used XMP workflow before, I would really appreciate some assistance with this. Is there a tutorial you can direct me to? I have a textured sphere shot for each camera so I can use that for scale/align etc but how can I import the camera settings to use for other scenes in RC?
    Thanks

    Jules
    0
    Comment actions Permalink
  • Avatar
    Thomas Van Damme
    julesmalcomson wrote:
    Alignment time is not really an issue for me, is there a way to increase the settings to get more reliable alignment?
    Or can someone explain the relevant parameters to achieve better success aligning images?


    I'm not working with a fixed array of cameras but I have the same question: alignment time is not really an issue, is there a set of parameters I can change which might make it so that RC takes longer to align but which assures highest probability of successful alignment?

    A sequence of steps for maximum success would also be good, ie:
    Step 1: use these settings, run align
    Step 2: delete everything but biggest chunk, change settings to this, run align again
    Step 3: repeat with these settings

    Any input or personal experiences welcome!
    0
    Comment actions Permalink
  • Avatar
    Wishgranter
    Hi julesmalcomson

    With the good aligned dataset use SELECT ALL from the ALIGNMENT ribbon and then use the Metadata (XMP) tool and export with settings in the screenshot, the XMPs will be saved next to the project images, COPY the XMPs to another folder with images which need to be transferred into the good alignment. Then just add them to the project and click Align. It will align the cameras and you will have all cameras aligned as with the good working dataset.

    XMP_export.png
    0
    Comment actions Permalink
  • Avatar
    callum
    Hi I've exported the XMP Data from a project with the correct alignments and have put them in a new image folder. After I press align with the new images it aligns correctly with all cameras in one component but there is no basic point cloud. After I process a normal detail mesh it is extremely broken. Does anyone know the solution for this?



    Thanks
    0
    Comment actions Permalink
  • Avatar
    Wishgranter
    Hi callum
    As I have written you in my email, it looks like the cameras are NOT the same... please follow the email instruction and visually compare the images...
    0
    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post