zhengthomastang / cal_pnp Goto Github PK

Hello, I want to ask if I can use a photo of different size as input. Currently it has some problems. How to adjust the code? Thank you!

hi.
I try to this code with a new opencv version 4.5.0.
But I can't do it at my pleasure.
So, Could you tell me what version I must try it.
Also, if we try version4.5.0, could you tell me points that should be revised of this code.

I will attend your reply.

Finally I run this program successfully and get the 3x3 homography matrix output. But in your CAM_CAL_IPL, the output is intrinsic camera matrix, the rotation matrix, the translation vector and the projection matrix.
Can 3x3 homography matrix transfer to camera parameters?

I used four 2d points and corresponding 3d points,but when I run the code,I get 0 reprojection error.
Any suggestion is greatly appreciated.

Thank you

Thanks for the great work @zhengthomastang .I tried calculating the 3d world coordinates from 2d pixel coordinates using the estimated homography using the provided code.
Here is the small snippet:
point=np.matrix(560,670,1) ## a homogeneous pixel coordinate of the image.
3dpoint=np.dot(point,homography.inv())## taking dot product of the 2d point and inverse of homography matrix.
scalar=3dpoint(0,2)##scale factor
3DX=3dpoint(0,0)/scalar ##Real world X coordinate
3DY=3dpoint(0,1)/scalar ##Real world Y coordinate

But I am getting weird values.Am I doing something wrong?
Any suggestion is appreciated.
Thanks