Extrinsic parameters, camera calibration

[ec01]

---

invancea a écrit :

The projection matrix is H (Homography) and this is what you get out of IMAQ Get Camera Parameters. You can construct camera matrix (M) out of Focal Length and Optical center. Matrix W = [R T] can be computed by using M and H.

---

What are the units of the projection matrix ?

To calculate the matrix W I have to multiply the inverse of M and H, I can construct the M-matrix in pixels or milimeters (by conversion using the cell-width) using the VI "get camera parameters", but I don't know how using the matrix H...

(the documentation "In-Depth Discussion" is complicated, it's written : fx = F/sx, with F in milimeters and sx in pixels/milimeters --> the fx value is in square-milimeters per pixels ?!!!?)

I'm a little lost in all this concepts, I hope my questions are clear enough (and my english too  )

[ec01]

Nobody has an idee to interprete the results of the calibration ?

[Klemen]

Hello,

maybe this can help you in some way:

http://forums.ni.com/t5/Machine-Vision/Camera-extrinsic-parameters-in-stereo-vision-calibration-and-...

Best regards,

K

[ec01]

Thank you Klemen,

I've reaad this thread but I didn't understand what is wrong in my procedure...

For instance in this piece of code I would'like to obtain the rotation matrix, but the matrix at the end isn't a rotation matrix (non-normalized vectors, the vectors aren't perpendicular...).

[leonardoMM]

I've the same problem

according to the help Cx and Cy need to be in [mm]

https://zone.ni.com/reference/en-XX/help/370281AC-01/nivisionconcepts/spatial_calibration_indepth/#h...

some one have found the soklution?

ciao

[leonardoMM]

Hi All

I've done some test

According to what is write on this Labview guide

http://www-cs.ccny.cuny.edu/~wolberg/capstone/opencv/LearningOpenCV.pdf

I think that the prjection matrix provided by the LV block "IMAQ Get Camera Parameters" is like this [R1 R2 T]

but need to manipulate with the SVD decomposition as describe on page 391.

With this trick det(R)=1 and R3=R1xR2

attached my VI with the value caming from my calibration

I will verify and validate the data as soon as possible.

Ciao

[Sekine108]

Hey leonardoMM,

Were you able to verify and validate the output of the IMAQ Get Camera Parameters? Did you produce expected results for your Rotation Matrix throughout your tests?

[xrtc211]

The problem of calculating the Extrinsics of the camera providing image points and world points refer to the documentation of Labview comparing with known Matlab function:

[R, t] = extrinsics(imagePoints, cam_params.WorldPoints, cam_params)

In Labview the vi Learn perspective calibration calculates matrix W=[r1 r2 t] (or not?)

The vi  Get Camera Parameters returns this projection matrix W. From what I found through the posts and documentation  I understand that projection matrix is the matrix W=[r1 r2 t].

From what people proposed we can decompose the matrix W and receive r1 r2 r3 vectors.

I am wondering why in Matlab the translation vector is a 3x1 which contains the offset in the Z direction of the checkerboard origin whereas in Labview it gives you a translation vector which does not contain Z.

If the vi Learn perspective calibration returns array H instead of W, where H=M*W  based on https://www.oreilly.com/library/view/learning-opencv-3/9781491937983/ page 407 the translation vector t=λ*(M^-1)*h3 where h3 is the 3x1 vector of array H.

I attach the array M of the intrinsic and the projection matrix. 

So the question is how to get the full translation vector t=[X Y Z 1] like Matlab function [R, t] = extrinsics(imagePoints, cam_params.WorldPoints, cam_params)

Thanks