CentralCamera.points2C#

static CentralCamera.points2C(P: ndarray, p: ndarray) → tuple[ndarray, float][source]#

Estimate camera matrix from data points

Parameters:

P (ndarray(3,N)) – calibration points in world coordinate frame
p (ndarray(2,N)) – calibration points in image plane

Returns:

camera calibration matrix and residual

Return type:

ndarray(3,4), float

Estimate the camera matrix \(\mat{C}\) determined by least squares from corresponding world P and image-plane p points. Corresponding points are represented by corresponding columns of P and p. Also returns the residual which is:

\[\max | \mat{C}\mat{P} - \mat{p} |\]

Example:

>>> from machinevisiontoolbox import CentralCamera, mkcube
>>> P = mkcube(0.2)
>>> camera_unknown = CentralCamera(f=0.015, rho=10e-6, imagesize=[1280, 1024], noise=0.05, seed=0)
>>> T_unknown = SE3.Trans(0.1, 0.2, 1.5) * SE3.RPY(0.1, 0.2, 0.3)
>>> p = camera_unknown.project_point(P, objpose=T_unknown)
>>> C, resid = CentralCamera.points2C(P, p)
>>> C
array([[ 852.8155, -233.0621,  633.5638,  740.0291],
       [ 222.9807,  990.0484,  294.836 ,  711.964 ],
       [  -0.131 ,    0.0655,    0.6488,    1.    ]])
>>> camera_unknown.C()
array([[1500.,    0.,  640.,    0.],
       [   0., 1500.,  512.,    0.],
       [   0.,    0.,    1.,    0.]])
>>> resid
np.float64(0.0376726171853079)

Note

This method assumes no lens distortion affecting the image plane coordinates.

References:

P. Corke, Robotics, Vision & Control for Python, Springer, 2023, Section 13.2.1.

Seealso:

C images2C decomposeC

CentralCamera.points2C#

This Page