reprojecting texture to mesh using 2 sets of object images

Hi there - I’m pretty much a newbie in the photogrammetry world …so bear with me if I’m asking questions that have been discussed elsewhere in this forum … and grateful for any helpful advise from the ‘Pros’ …

I am planning to use CR for my full body photogrammetry scanning booth (currently under construction) and create models from humans for 3D prints of full color mini-replicas (human figurines).

In order to achieve the best possible model quality for 3D print I am planning to run the capturing process (based on 144 Raspberry Pi 3 and Camera Modules v. 2.1 , 8 Mpx) in 2 stages.

Capture 1: 144 pics of the object using 6500k LED light
Capture 2: 144 pics with slightly dimmed LED lights and random patterns projected onto the object (using high Lumen data projectors) in order to improve mesh quality for homogeneous object surface (e.g. person wears black suit, shiny suit, etc)

The 2-stage process will occur in a split of second… and the object can be considered stable / unmoved.

Now I would like to use CR to compute a clean mesh from capture set 2 ( with projection patterns).
As for the texture I would obviously prefer to process image set 1 (without projection patterns).
Lastly I would like to project texture from image set 1 onto mesh from image set 2 to create an exportable obj-model file that can be shared for potential cleanup and final 3D print stage.

I hope you’re getting my point? …and the exam question is:

Can I do this with CR and if yes - HOW???

Thanks in advance everyone!

PS: I know Photoscan from Agisoft has such a feature …but I love CR’s speed and model quality…hence I really would like to give CR a go…before being forced to use Agisoft…:slight_smile:

Hi Sabry
Yes, it is possible, but if you want to load the scanner fully, then I recommend you to get the CLI version where this is solved very nicely.
With CLI you can reconstruct models fully automatically, properly oriented and scaled, and also reuse it for reconstruction of hard to align projects.

If you use the ordinary PROMO version, you can do it this way:
D1 -> images with patterns
D2 -> images for texturing

  1. when you have set everything and reconstructed a model from D1, then SAVE the project and exit or create a new project ( so that you do not have the actual dataset open in RC )
  2. move out D1 images from its actual folder somewhere else ( proper planning of data workflows )
  3. Open the same project again, it will ask you where are the MISSING images ( D1 ) and you just point it to the location of D2 images, then just use UNWRAP and texture for projecting proper textures.

Many thanks for your simple and pragmatic advise…the CLI version would be indeed very useful …as we may expect heavier scanning volume down the road…but being a startup we need to manage cost too for now…and test the market before upgrading our technology base (incl. software licensing)…I suggest you guys have been through a similar phase when CR was founded… :slight_smile:

I will test your advise and get back with any questions as needed…

Hey Wishgranter, thanks very much for your advise and your approach works well. However, I have been experiencing a problem that the model is not loaded after I saved and reopened the project.
I also tried to save another project and then reload the problem happened again. Just wondering have you ever experienced similar issue?
Thanks very much!

Regards,

Wishgranter wrote:

Hi Sabry
Yes, it is possible, but if you want to load the scanner fully, then I recommend you to get the CLI version where this is solved very nicely.
With CLI you can reconstruct models fully automatically, properly oriented and scaled, and also reuse it for reconstruction of hard to align projects.

If you use the ordinary PROMO version, you can do it this way:
D1 -> images with patterns
D2 -> images for texturing

  1. when you have set everything and reconstructed a model from D1, then SAVE the project and exit or create a new project ( so that you do not have the actual dataset open in RC )
  2. move out D1 images from its actual folder somewhere else ( proper planning of data workflows )
  3. Open the same project again, it will ask you where are the MISSING images ( D1 ) and you just point it to the location of D2 images, then just use UNWRAP and texture for projecting proper textures.

Hi Tantan Xie

have you selected proper COMPONENT with that reconstructed model ?

Screenshot 2016-02-03 08.28.52.png

I had an issue with something similar where I needed to change the source images. If you name them something similar, say
image_0001_scan
image_0001_project

or whatever, you can open the project file and search replace the image lines there to your new images.

That is exactly where my problem is. I selected the proper component and now everything is working. Much appreciated!

Wishgranter wrote:

Hi Tantan Xie

have you selected proper COMPONENT with that reconstructed model ?

Screenshot 2016-02-03 08.28.52.png

Which projector did you choose? I am also trying to get a similar setup.