Question

I have a rotated (3D) rectangle object in real-world and I'm trying to re-project this object into the image.

Suppose we have the following parameters:

  • The four corners of the rectangle in real-world and the image [(xw1, yw1,zw1) , (xw2, yw2,zw2) , (xw3, yw3,zw3) , (xw4, yw4,zw4) ] ; [(x1, y1) , (x2, y2) , (x3, y3) , (x4, y4) ].
  • The distance between the camera and the object (e.g. D).
  • Intrinsic camera parameters (focal length and principle point).
  • Extrinsic camera parameters (the rotation matrix R (3x3) and the translation matrix t (3x1)).
  • The 3D surface normal vector (<v1; v2; v3>) of the rectangle object with respect to the camera.
  • Check this image: image

How can we re-project the rectangle from the "real world" to the image using the given information above. So, the object in image must look like the real world (oriented).

Was it helpful?

Solution

You can compute a projective transformation from the images of four points. So the first point you mentioned, the coordinates of the rectangle in world and image coordinates, is sufficient.

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top