¤ 1. Run detect,
Mark's corner detector
|
| INPUT: |
an image of the calibration aid |
| OUTPUT: |
a DAT file detailing corners detected in the image and their strengths |
| See sample output |
|
¤ 2. Run analyze,
Mark's data extrapolator, which takes detect's output
and figures out equivalent data for infinitesimal window sizes, attempting to place corners with
sub-pixel accuracy (this is a large task, seeing as how images don't have sub-pixel accuracy!)
|
| INPUT: |
eight data files, output from detect,
giving information for all eight images of the calibration aid
a flag telling whether we want information for the xy-plane or the z-axis |
| OUTPUT: |
a DAT file detailing all corners from the images in either the
xy-plane or the z-axis |
| See sample output |
|
¤ 3. call calibrate, my new customizable version of Zhang's algorithm, which performs whatever manipulations I want on the input data (corner coordinates), calibrates the camera and spews out information on the camera's properties and the homographies (coordinate systems) of all the images used for input
|
| INPUT: |
the same as easycalib: DAT files with corner locations and a model file
|
| OUTPUT: |
a calibration file with information on the intrinsic properties of the camera and information about the locations and orientations of all the images that produced the input data files
|
| See sample output |
|
¤ 5. call zerror3d,
Mark's final statistic generator, to display output on the 2-D and 3-D statistics of the error
in the current routine (of course, only the error not generated by Zhang's calibrator matters
to us)
|
|
| INPUT: |
eight input files for eight images, containing only detected corner coordinates,
and the one output file produced by Zhang's program, containing either 2-D or 3-D calibration
information |
| OUTPUT: |
statistics on 2-D and 3-D average errors, given known point coordinates, of
corner coordinates generated from a camera with the internal parameters given by Zhang's
program's output |
| See sample output |