Application of pinhole distortion values and ftheta bw-poly

Dear nvidia-team,

we are trying to refine the transformation from camera pixel coordinates to world coordinates. About this, we have two questions concerning entries in the rig-files obtained by calibration:

  • When calibrating a single 60° camera (or at least that is what it seems to us), it may be that the calibration result uses the pinhole camera model. There are distortion values given there, and we wonder how to use them, exactly. Are these the coefficients of the even-order polynomial model (the k_x in the formula below)?
    (x/y]_undist = [x/y]_dist * (1 + k_0 * r² + k_1 * r^4 + …)

  • Other cameras yield the so-called f_theta camera model (we figured it is a way to simulate an undistorted image as if it were from an f_theta lens via software). There another set of entries called bw_poly are given. How are these values used, what kind of polynomial are we looking at here? Are we calculating in pixels or meters, in degree or radians?

  • Where in the calculation pipeline are these rectifications applied? right on the pixel values? after rotation and translation?

Many thanks!

Dear @josua.zscheile,
Is the discussion in Parameter support in pinhole camera model helps?

Dear @SivaRamaKrishnaNV,

thanks, that should help for the pinhole model, it seems that it is as we expected.

For f_theta, the entries remain unclear wrt. how they are used.

Any more input on how the polynomial for the f theta model works?

Dear @josua.zscheile,
Have you checked Ftheta backward polynomials details