Forum

Author Topic: Batch processing - Setting scale and orientation and Reconstruction Regions?  (Read 9586 times)

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
Hey guys,

I've been testing out batch processing in PS and so far it's been amazing. I just have a few questions on alignment, scale and reconstruction regions.

So first off, my alignment is always incorrect inside of PS, how would I go about fixing this? and how would I automate this with Batch processing?

http://i.imgur.com/eA68HdP.png

In regards to scaling;  I've used Control points on my neutral scan and set the scale bar to 4.0 and hit update, this sets up the correct scale of 4cm between the control points but doesn't PS work in meters? and if so, shouldn't the correct value be 0.04?

Also in regards to scaling, how would I force all the chunks to follow the same scaling as Chunk01(Neutral expression Scan)

Lastly is there anyway to change the reconstruction region in batch processing? Currently it cuts out the ruler on the left hand side which I'd like to keep in for reference.

Any help is much appreciated,

Rashed

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
I should note: I'm currently using a 30 day trial of Photoscan Professional

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15413
    • View Profile
Hello RashedAlmet,

The default coordinate system in PhotoScan is in meters, however, most of the 3D mesh formats do not save the coordinate system definition, so if you are treating the units of the model as cm in the post processing software, you'll get 4 cm instead of 4 meters.

To apply the same scale to all the chunks in the project, you can align the chunks, for example, using Camera Based approach, providing that the labeling convention for the cameras in the separate chunks is the same.

Batch processing currently doesn't include the region resize instrument and only Reet Region option is available. For more customized batch processing you may consider using Python scripting. Or use some basic scripts prior to the processing, like applying the same region box to all the chunks, copying it from the first one.
I think you can check the following posts from Infinite Realities blog (not sure, though, that the scripts are up-to-date:
http://ir-ltd.net/delivering-aligned-and-scaled-photoscan-outputs/
http://ir-ltd.net/its-your-birthday/
Best regards,
Alexey Pasumansky,
Agisoft LLC

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
Hey Alexey and thank you for the quick and detailed reply.


Will the 'Align Chunks' command; align all chunks to the same scale as Chunk01? (So I can set proper scaling using Chunk01 then that get's copied to all the other chunks)


Also how do I properly set alignment inside of PS? Every scan I've done so far has been orientated at a weird angle (see image below)
http://i.imgur.com/Tw1kChB.png





Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15413
    • View Profile
Hello Rashed,

In the Align Chunks dialog you can select the Reference chunk by double-clicking (it will be marked with the bold font) - the orientation and scale of this chunk wouldn't change, while for other chunks their positions, orientation and scale will be adjusted accordingly.

If you need to modify the orientation of the model relatively to the coordinate system axis, you can use Rotate Object tool (on the Toolbar) or input the coordinate information for the markers in scene or the camera locations that will define both scale and orientation of the model in the coordinate system.
Best regards,
Alexey Pasumansky,
Agisoft LLC

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
Hey Alexey,

I've run through camera alignment but I'm getting these errors in the console:
http://i.imgur.com/DmCaa01.png

Is this because Chunk 2 is using 80 Cameras, whereas Chunk01 is using 88?



Setting coordinate information or camera locations is new to me. Is there any tutorials or documents you can recommend to show me how to do this?

My current Point setup:
http://i.imgur.com/Fd9NnUR.png

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15413
    • View Profile
Hello Rashed,

Chunk alignment options are described in the Chunks section of PhotoScan Pro manual.

Camera Based approach is applicable ONLY if chunks contain the cameras with identical labels related to the same real world locations. Usually this approach can be used for the fixed camera rigs in studios.
Best regards,
Alexey Pasumansky,
Agisoft LLC

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
Hey alexey,

This is something we have. All 90 cameras are attached to a fixed rig. Would I find information on the camera approach inside the pro manual too?

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15413
    • View Profile
Hello Rashed,

Do you have identical labels for the corresponding camera positions in all the chunks that are being aligned?

Also I think you've misread the console output, "error" is related to the internal operations and can be treated as accuracy of the results. You should check for the chunks, if the [R] mark appears next to the chunk labels, if so, then they are "referenced" and you should be able to display all of them at once using Show Aligned button on the Toolbar.
Best regards,
Alexey Pasumansky,
Agisoft LLC

RashedAlmet

  • Newbie
  • *
  • Posts: 13
    • View Profile
Sorted it :) DXO was saving out the second batch with a different name.

Everything is nicely aligned now; just slightly out of place (because of the actors head movement I believe)



Thanks for your help Alexey!
« Last Edit: July 03, 2017, 02:24:51 PM by RashedAlmet »