I am calibrating the camera and it seems to work fine but I would like to know what are good values for the average and maximal residue? Also, what impact does a few missing dots (using show grid points when using grid of dots) or squares (using “Show features” with checkerboard) have to the calibration result?
Using either grid of dots or checkerboard have the same purpose, to translate camera pixels to millimeter and to establish a camera origin.
If a couple of dots/squares are missed during calibration the camera will not be calibrated in those areas where the dots/squares are not found. Dots or squares not found are usually caused by poor lighting (i.e. glare, shadows, poor focus, iris etc.). The goal should be to always find all dots/squares, it can however be very hard with poor lighting.
The max and average residual indicates how good the outcome of the calibration is. A high residual indicates poor results. An average residual of less than 0.3 can be considered ok, even though it usually is around 0.1. The max residual indicates if there is one or more dots/squares that has a significant offset (compared to where it is supposed to be).
Hi,
I’ve done the dots calibration, but i have a question: i have been able to find all the dots in the calibration paper and the max residual is about 0.18 (the average residual about 0.05), but sometimes when i leave some dots undetected (like 5 or 6 dots) , the max and average residual becomes even smaller (0.09 and 0.04). So which calibration should i choose if i want to have a very precise camera calibration (all dots or the lower the residual the better?) ?
Another thing: I’ve done the checkerboard calibration too. It is supposed to be for an even more precise calibration, but once i did it, i got a max residual of 3.91 and an average residual of 0.54. So i inmediately rejected this type of calibration. I’m not using the CameraCalibPaperCheck20A3.pdf (which is not aligned properly). What can i do if i need a very very precise calibration? stay with the dots?
Thanks for your attention,
Rachid
The dots that you are not finding are probably the ones that are in distorted in the edges of the view. So thats why you get better values without them than with them.
The values that you are comparing are not the same for the dot calibration and the checkerboard calibration. If you have a perfect camera, lens and perfectly aligned view of the frame you can use the dot calibration. However the checkerboard calibration is superior in all the other cases.