Given the extrinsic (including center of projection) and intrinsic parameters of a stereo pair, and enough correspondences (say 10-11), is it possible to triangulate the exact 3D position of the correspondences?
Some websites say it is only possible to determine the relative orientation of the without control points. These websites say it is a relative triangulation -- correct up to a similarity transform. But that doesn't make sense if you know where the cameras are located in world reference frame. What is the correct answer?
And is this related to projective ambiguity at all? These terms seem related.
If you know the intrinsics of both cameras and the rotation and translation between them, then you can certainly triangulate the 3D positions of the correspondences in world units. How exact those positions are depends on the resolution of your cameras, the accuracy of the camera parameters, and the accuracy of the correspondences.
If you don't know the rotation and translation between the two cameras, then you get an up-to-scale reconstruction.
If you also don't know the intrinsics, then you get projective ambiguity.