Stereo camera calibration

asked 2017-07-26 18:07:16 -0500

BhanuKiran gravatar image

updated 2017-07-26 18:08:38 -0500

I ran stereo calibration

RMS error = 2.5624 average epipolar err = 2.79867

I am not sure, if these values make sense.

Left Image

image description(/upfiles/15011100969575928.jpg)

Right Image

image description

Rectified image

image description

edit retag flag offensive close merge delete



0.5 or smaller would be a nice calibration error, rinse and repeat.

how many images did you take ? just those 2 ? you'll need like 20-30 pairs, the whole field of view should be covered, especially the corners of your image.

make sure, your paper chessboard is absolutely plan, no wrinkles or such, any millimeter counts !

berak gravatar imageberak ( 2017-07-26 23:10:40 -0500 )edit

Thank you for the comment. Previously, i was working with only 14 image pair. Now i tried with 29. I see the epipolar lines matching nicely unlike above image. But RMS error = 1.56 average epipolar error = 2.6 . still remain higher.

cv::CommandLineParser parser(argc, argv, "{w|9|}{h|6|}{s|20|}{nr||}{help||}{@input|stereo_calib.xml|}");
float squareSize = parser.get<float>("s");

My square size is 20mm. The argument for variable S should in mm right.

BhanuKiran gravatar imageBhanuKiran ( 2017-07-27 15:45:50 -0500 )edit