Reference frame of transformation obtained from estworldpose

Hi,

I have a set of 3D points (x,y,z) from a depth map in the Camera Frame aftter proyecting them using the camera intrinsics. Then, I got a set of 2D features (u,v) of that same object from a different unknown Camera pose. I am trying to recover the transformation between the Unknown Camera Pose and the 3D points.

I am trying to use PnP to find the transformation between 3D points and 2D points. However, I am confused on this transformation based on the documentation: "Estimate camera pose from 3-D to 2-D point correspondences" in estworldpose. I am trying to wrap around my head if the final transformation TF is from World to Camera or from Camera to World. It might be a dumb question, but semantics can change the meaning completely. Will it be:

or

Appreciate the help :)

Answer 1

[Vinayak]

Hi Daniel,

I understand that you wish to clarify whether “estworldpose” returns object from World to Camera or Camera to World perspective.

When using functions such as “estworldpose” or other Perspective-n-Point (PnP) algorithms, the goal is to estimate the camera's pose based on correspondences between 3-D world points and 2-D image points. Thus, the resulting transformation typically represents the camera's position and orientation within the world coordinate frame. Therefore, the transformation matrix “TF” you obtain will map points from the world frame to the camera frame.

To transform a point from the world coordinate system to the camera coordinate system, you would use the “TF” matrix to multiply the point's homogeneous coordinates.

If you need to perform the opposite—transform points from the camera frame to the world frame—you would use the inverse of the “TF” matrix.

Hope this helps. Please feel free to reach out for further clarifications.