ethz-asl / kalibr

The Kalibr visual-inertial calibration toolbox
Other
4.43k stars 1.4k forks source link

Significant Discrepancy in xi Parameter between Kalibr's Omni-Radtan Model and CamOdoCal's MEI Model #706

Closed qpoisson closed 2 months ago

qpoisson commented 3 months ago

I have been using kalibr_calibrate_cameras with option omni-radtan to calibrate a single fisheye camera. The calibration process outputs an xi parameter of approximately 1.3. However, when I use CamOdoCal with the MEI model for the same calibration data, I obtain an xi parameter of around 0.9.

My understanding was that both tools should provide similar results, as they are both used for calibrating fisheye cameras with MEI model. Could you help clarify if these two models are equivalent and why there might be such a significant difference in the xi parameters obtained from each tool?

Thank you for your assistance.