New calibration target design, code to scan it with....

Hi folks,

Dunno if anyone else here plays with camera calibration, I thought I’d let people know that I’ve written a better calibration target scanner than the one in OpenCV (cvFindChessboardCorners). I’ve been frustrated for a while with the fact that it requires that the entire checkerboard be visible in the image. I also required for a project that the calibration target had a fixed origin, so that if it were flipped 180, I could tell. This code remedies both of those problems.

So the code will scan any part of the checkerboard that is visible in the image, and always knows exactly where it is on the checkerboard, since each square carries an indentifying fiducial. It’s also quite quick: ~.05 sec to scan a 640x480 image on a very modest machine.

I wrote a blog post explaining it here.

Here are some sample target scans:

http://i126.photobucket.com/albums/p101/kefex/scanned-1.jpg
http://i126.photobucket.com/albums/p101/kefex/scanned-2.jpg
http://i126.photobucket.com/albums/p101/kefex/scanned-3.jpg

And the code is here. The code includes a few test images, a pdf of a sample target, and a python script for generating your own custom targets.

Cheers,

Kevin Atkinson

P.S. please drop me a line if you find the code useful! kevin.atkinson at gmail.com