Ftheta backward polynomials details

Autonomous Vehicles DRIVE AGX Xavier DRIVE AGX Xavier General
1

Hi,

Descriptions of the ftheta model parameter provided in the post1 and post2 show that ftheta model of the driveworks resembles the OCAM model as ftheta model does not require an affine mapping process. I’ve tried to follow the documentation and the image undistortion results were completely different from what I’ve expected.
I have applied direct transform with backward polynomial and also estimated forward polynomial(invpol) and computed again and got the same different results.

Therefore the question is

  • Can you describe more details about the backward polynomials of ftheta model?
    I think the ftheta model is not just the simplification of OCAM model.
  • Also if possible please share references or documents about the ftheta model other than
    ‘A Generic Camera Model and Calibration Method for Conventional Wide-Angle, and Fish-Eye Lenses’ .
4

Hi @junghee.kim ,

We will check internally and update you. Thanks.

5

The whitepaper we shared explains exactly the model and how to compute pixel from rays and vice versa. Yes, f-theta is not using affine transformation. Model parameters of ocam model cannot be used straight in f-theta model. For example pixel2ray mapping in OCAM model is computing Z using the polynomial, see for example: Fisheye Calibration Basics - MATLAB & Simulink - MathWorks Deutschland
While in our model the polynomial describes the relationship of the distance to the principal point to ray angle to the optical axis, see Section 2 (Backward polynomial) of the f-theta whitepaper.
These are totally different quantities and hence these are different polynomials.

1 Like