Object 3D world coordinates from multiple images

I Want to find Object 3D world coordinates (called Photogrammetry) using coded targets

Using single camera, multiple images

Known

• Object 2D image coordinates (x,y) == coded targets
• Camera intrinsics, extrinsics parameters (using Calibration Checkerboard)
• Camera world pose from each images ( Image 1 : R,t / Image 2 : R,t ••• ) - Extracted R,tvec by camera calibration method

Unknown : Object 3D coordinates (x,y,z) From real scale

i already use worldPoints = triangulate(matchedPoints1,matchedPoints2,camProjection1,camProjection2)

this func and get worldpoints from two images (Using a known distance of length like Ruler)

and the precision was not bad either.

Multiple images, I know TriangulateMultiview and bundleAdjustment

But this func impossible to calculate 3D-coordinates considering the real 'Scale'

1. How do I get 3D - coordinates that are scaled from multiple images?
2. Sfm 3d reconstruction can solve this questions?

Answer 1

[Qu Cao]

You will need a stereo camera to give you the actual dimension of 3-D objects.

Alternatively,if you know the size of an object in the triangulated 3-D points, you can estimate the scale. See this section of the Structure From Motion From Two Views example.

[sksk]

You will need a stereo camera to give you the actual dimension of 3-D objects.

• you mean, single camera using stereo visions and stereo cameras are different?

Alternatively,if you know the size of an object in the triangulated 3-D points, you can estimate the scale.

• yes i did it with known the size of an object (Like Ruler)

See this section of the Structure From Motion From Two Views example.

• it takes only two views examples and i saw Structure from motion multiple Views on Matlab but i think it can't estimate real scale of object,,, cuz i need both accuracy and precision 3d-coordinates