User instructions of the camera calibration toolbox for MATLAB
1999, Hynek Bakstein - hynek@terezka.ufa.cas.cz,
Radim Halir - halir@ms.mff.cuni.cz
The directory structure of the toolbox distribution is organized as follows:
/samples - contains sample data
/samples/images - sample images form real data experiment
/samples/coords - coordinates of reference points detected in sample images
/m - MATLAB files implementing the calibration method
/solvopt - optimization routine used by the method's implementation
There are also several files in the root directory. The 'README' file
contains user instructions and the 'demo.m' file contains a MATLAB
function 'demo', which implements a simple demonstration of presented
method.
The main function implementing the proposed method are commented. All
the functions contain help information. Each file contains one
function, and has the same name with the .m extension.
The calibration is performed by the function 'calibr8'. It has a
variable number of arguments, but first three have always be passed.
These are the initial estimate of the intrinsic parameters, the maximum
number of iterations and the desired precision level. The first
parameter, the initial estimate of intrinsic parameters, is represented
by a vector of the following parameters: the focal length, the u0 and
v0 coordinates of center of CCD chip linear distortion coefficients b1
and b2. The second parameter, the maximum number of iterations, is and
integer number larger than zero and the third one, the precision level,
is a positive floating point number.
Next parameters specify the measured data. For each plane is one
variable Their number is optional. At least two planes are needed for
the method to succeed. The coordinates are in form of a n by 5 matrix,
where each row specifies the pairs of coordinates of control points and
their images and n is a number of such a pairs in the plane. The first
three columns of the matrix represent the xi, yi and zi coordinate of
the reference point. The fourth and the fifth column represents the ui
and vi coordinate of the appropriate detected image of such a point.
The function returns four parameters. The first one is a vector of
estimated intrinsic parameters having the same form as the initial
estimate parameter of the function. Next parameters are the estimated
translation and rotation matrices of the camera and errors in each
plane.
The function 'demo' implements simple demonstration of the proposed
method. First it adds all necessary directories into the path, where
MATLAB searches for functions. Then it loads the demo data located in
the 'samples/coords' directory and runs the calibration by invoking the
'calibr8' function with the demo data. After the calibration is
finished, the results are being displayed along with graphs of the
resulting errors.