# Extrinsic.yml meanings

Hi, all,

I have run the stereo calibration example and gotten both intrinsic and extrinsic parameters of my stereo camera. I set the squareSize= 108.f since my chess board using 108 millimeter 8X6 squares. But I no ideal what correct unit the square size should be. And finally I got the rotation and translation matrix.

R:[9.9971067186655571e-001, 1.0255094565610759e-002,-2.1757885732999915e-002,
-1.0256469002886417e-002,9.9994739986591286e-001, 4.8425128267447104e-005,
2.1757237869562643e-002, 1.7474796307336661e-004,9.9976326801070092e-001]

T: [ **-2.1943586431950470e+002**, 3.0918889737998665e+000,-3.3488303392429639e+000 ]


I measured and my basic line is around 213.05 mm. I think the first element of T, -2.1943586431950470e+002, represent the basic line length.Is it right? Could you guys share me reference showing the meanings of intrinsic and extrinsic parameters? Thanks!

edit retag close merge delete

hello, I am workin on same problem, could I ask you how you get these parameters from stereo vision?

( 2013-05-20 09:59:14 -0500 )edit

Sort by » oldest newest most voted

The extrinsic parameters are:

• Rotation: the rotation matrix (3x3) from the origin of the world (usually the inner corner of the chessboard when you specify (0,0,0) for the first 3D points during the calibration process)
• Translation: the translation from the first inner corner of the chessboard.

The intrinsic parameters are:

• The focal (in pixels) on the diagonal (fx, fy)
• the center point of projection (cx,cy) on the last column
• a coefficient of distortion (k) [ fx k cx 0 fy cy 0 0 1 ]
more

Hi there!

Don't worry about the units of the square sizes, the results will be in whatever units you provide the square size. For instance, if you provide the square size in millimeters then the results will be in millimiters, if you provide the size in Kilometers the the results will be in Kilometers. As long as you provide an accurate estimation of the square size your results will be good.

About the meaning of the rotation and translation matrix, it is simple: R and T represent how you should translate and rotate the right camera, in 3D space, to match the image plane of the left camera. Therefore, if your base line is around 213.05mm, the first value of vector T makes perfect sense, as it represents how much you should move the right camera along the X axis to match the position of the left camera.

Learning OpenCV: Learning Computer Vision with the OpenCV Library. Chapter 12: Projection and 3D vision.

more

Your information really help me. Thank you very much! By the way, I have no ideal about the stereo camera calibration flag, CV_CALIB_ZERO_DISPARITY. Dose it mean translate and rotate the left camera to match the image plane of the left camera with the same principal point Cx and Cy?

( 2013-02-21 16:38:07 -0500 )edit

Official site

GitHub

Wiki

Documentation