How to do a manual calculation of worldpoints from image points with Matlab Computer Vision data

I am Trying to get a grasp on computer vision. So far I have managed to calibrate a singel camera using Matlab Camera Calibrator, and done a planar object measurement that was incredibly accurate using an adapted example given on this site. Matlab calculates worldpoints automatically, but I want to learn how to do It manually using intrinsics and extrinsics. Can anyone here help explain to me how to do It? I feel I am pretty close to accomplishing my goal, but I am stuck at the moment.

From Matlabs Camera Calibrator I get these intrinsic og extrinsic parameters

intrisic matrix = [3344.2 0 0 ; -0.6732 3330.6 0 ; 2141.6 1475 1];

rotational matrix = [0.9996 0.0267 -0.00058362 ; -0.0267 0.9996 -0.0049 ; 0.00045327 0.0049 1]

translational vector = [-180.3688 57.61 1427.3 ]'

Using the cameraParams, rotational matrix and translation vector i get the results I want with the pointsToWorld function, but when trying to calculate this manually I don't. I have tried setting up the full camera matrix and multiply this with the worldpoints to get image points, and the invers.

given the Image points [1582 957] the pointsToWorld functions give me these worldpoints [-66.1856 -278.1794]

I would really appreciate It If someone could please spare some time to explain this to me

Answer 1

[Tohru Kikawada]

The following document page for pointsToWorld method would be helpful:

https://www.mathworks.com/help/releases/R2016b/vision/ref/cameraparameters.pointstoworld.html

You can also see actually how pointsToWorld calculates world points with the following command:

>> edit vision.internal.calibration.CameraParametersImpl

Hope this helps

[HT]

Thank you, this helps alot :-)

[Mingming Zhao]

Really? Did you get what you have expected? This function gives us nothing more than the final world coordinates(where z=0).