2D image and 3D model point correspondence OR Camera Pose estimation with x1 3D model and x1 photo.

[lNatxii]

Hello.

I am developing an application where everything is based on a 3D model which is actually a scan of an object. The scan is already done, and I don't need to modify it. The complex point comes when I want to be able to take N photos of the object and be able to map those photos as textures on the object. Let me explain:

Imagine I have a scan of a car. One day I want to take a thermal photo of that car, and be able to use it to preview, in a certain region of the car (the one that fits in the 2D photo) the texture of the thermal photo, but wrapped on the model.

There are three options for this:

• Is there any way that, having a 3D model textured with the rgb albedo of the object, the photo can be analysed and the texture can be mapped to the region of the object it is looking at?

• If not, is there any way to assign N control points (relate 2D points of the image and 3D points) to paste all the texture in the 3D model taking into account those imposed control points?

• Finally, if not, is there any way, having the 3D model and the photo, to know in which position and rotation relative to the object was the camera? Because in that case I do have a solution in mind.

Thank you.

Translated with DeepL.com (free version)

[natowi]

Sorry, this is not possible with Meshroom as it is