Best workflow for setting scale & orientation?

Development Asset Creation
,
1

I am working on body scans and would like to the know the best workflow to

  • set scale
  • set orientation (coordinate plane)
  • remove floor

I used a script in photoscan that automated this procedure with the use of target markers. Is there anything similar to this in Capturing reality?
http://www.pi3dscan.com/index.php/instr … projection

Thank you.

2

Hi Sung,

is your rig fixed? In RealityCapture you can export XMP files - classical extended meta file - extended with our tags with cameras parameters. XMP’s are associated with images, e.g. if you have

image1.jpg
image1.xmp

then the system will automatically load image1.xmp with image1.jpg and it will set camera parameters.
You can actually define whether the camera pose is fixed, should be optimized or it is simply not defined (simply delete the XMP file).

What is it good for? If you have a rig with almost fixed camera poses (not necessary perfect),

  1. take some well textured objects,
  2. capture images,
  3. define scale, define ground plane and
  4. align images.
  5. then export XMP files. You can optionally delete XMP files corresponding to cameras which you typically adjust.

Reusing data:

  1. take pictures of some other model and name images the same way like in the previous scenario,
  2. copy previous XMP files to the folder with images
  3. and align scene. It will take just a fraction of time, since almost all is already calculated.

scene will come in the same coordinates as the previous one. You can load reconstruction box, clip scene, calculate model,…
You can do this all using CLI version automatically.

3

Hi Sung

first have a look at how to work with ContolPoints
https://www.youtube.com/watch?v=d8naLEtLqDY

then set it to proper positions with your data and REalign the scene afterwards so that the actual scale is applied to the scene.
(or you can set it just after image import in RC if you feel it is more handy)

  • set orientation (coordinate plane)
    if you have coordinates of the CPs, you can IMPORT them via WORKFLOW -> GROUND CONTROL and place them on images, at least 3 GCP projections (1 CP-GCP on at least 3 images)

  • remove floor
    I am not sure what you mean by this, but if you need to remove groud floor, set the RECONSTRUCTION region and use the FILTERING tool and filter out what you do not need.

BUT if you use it for full-body and DO NOT change cameras (position, orientation + FILENAMES), then it will be even easier to use XMP worklow -> set everything at once, export XMPs and copy it to a new dataset (image folder) and then just align and reconstruct, it will keep scale, orientation etc.

4

Is there a way to define the ground plane in a way similar to how you define distance?

I’ve seen other photogrammetry software that allows you to specify that 3 control points are on the one horizontal ground plane, and thus it aligns the ground plane in relation to these points.

Currently I’m struggling to do it by hand using the “Define Ground Plane” gizmo as I have a large model that contains a canal so it needs to be very precise.

Help would be much appreciated on this.

5

Hi cillinfarrell
Yes, you can use 3 GCPs ( ground control points ) to get the proper orientation, scale and size.
You need to have XYZ coordinates for all 3 points, best if they are in *.txtr file for easier import etc.