changes.txt

tsai经典标定程序 MATLAB语言

----------------------------------------------------------------------------

Version 3.0b3  28-Oct-95

1) Added a check in coplanar_extrinsic_parameter_estimation () to choose
   the correct rotation matrix solution from the two possible solutions.
   This change affects only the coplanar extrinsic calibration routine.
   The coplanar and non-coplanar routines have had this check since
   release 2.0.

   Bug tracked down by Pete Rander <[e-mail]>

2) Added a check in undistorted_to_distorted_sensor_coord () for the rare
   situation where an undistorted sensor point must be mapped to the
   maximum barrel distortion radius.  The code will now map the point
   properly and print a warning when this happens.

3) Reworked and extended the FAQ file.

4) Changed the parameter declarations in routines

        cc_compute_exact_f_and_Tz_error ()
        cc_five_parm_optimization_with_late_distortion_removal_error ()
        cc_five_parm_optimization_with_early_distortion_removal_error ()
        cc_nic_optimization_error ()
        cc_full_optimization_error ()
        ncc_compute_exact_f_and_Tz_error ()
        ncc_nic_optimization_error ()
        ncc_full_optimization_error ()

   to match the declarations used in the MINPACK routine they're called by.
   No functional impact.


----------------------------------------------------------------------------

Version 3.0b2  15-Jul-95

1) Use the calloc () definition in the standard ANSI C stdlib.h
   header file rather than trying to define it locally.
   The code now works under DOS Borland C++ 4.5.

   Fix found by Markus Menke <[e-mail]>

2) Changed the variable declarations in routines

        cc_compute_exact_f_and_Tz ()
        cc_five_parm_optimization_with_late_distortion_removal ()
        cc_five_parm_optimization_with_early_distortion_removal ()
        cc_nic_optimization ()
        cc_full_optimization ()
        ncc_compute_exact_f_and_Tz ()
        ncc_nic_optimization ()
        ncc_full_optimization ()

   to match the parameter declarations used in the MINPACK routine they call.
   No functional impact.


----------------------------------------------------------------------------

Version 3.0b1  21-May-95

1) Modified routines

	cc_compute_exact_f_and_Tz_error ()
	cc_five_parm_optimization_with_late_distortion_removal_error ()
	cc_five_parm_optimization_with_early_distortion_removal_error ()
	cc_nic_optimization_error ()
	cc_full_optimization_error ()
	ncc_compute_exact_f_and_Tz_error ()
	ncc_nic_optimization_error ()
	ncc_full_optimization_error ()

   so that they return error values to the optimization routine rather than
   the squared error values.  The optimization routine "lmdif" squares the
   returned error internally during its calculations.  Before this change
   "lmdif" was essentially optimizing the square of the squared error
   (i.e. error^4).

2) Translated the MINPACK non-linear optimization routines from FORTRAN to C.
   The source code is now 100% C.

3) Re-wrote the matrix manipulation routines to use malloc instead of calloc.
   Malloc caused bus errors due to memory alignment problems on some platforms.

4) Replaced typedef enum { FALSE, TRUE } bool; in matrix code with #defines.
   The enumerated types didn't necessarily map correctly to the values required
   for logical checks like if (boolean_variable) then.

   Bug reported by Ron Steriti <[e-mail]>

5) Re-wrote the statistical calculations in routines

	distorted_image_plane_error_stats ()
	undistorted_image_plane_error_stats ()
	object_space_error_stats ()

   to improve their efficiency.

6) Fixed CBRT (cubed root) routine so it handles negative arguments properly.
  
7) Split up source code and renamed files for PC/DOS compatibility.

8) Created makefiles for Unix C and DOS Borland C++ 4.5.
   
9) Created a Frequently Asked Questions (FAQ) file. 


----------------------------------------------------------------------------

Version 2.1  04-Jun-94

1) Added code and a compile-time switch for using *public domain* MINPACK
   non-linear optimization routines.  Default compile uses commercial IMSL
   optimization routines.

   Code donated by Torfi Thorhallsson <[e-mail]>

2) Added a missing division by cp.sx in cc_nic_optimization_error () and
   cc_full_optimization_error ().  Bug reduced the accuracy of full coplanar
   calibration and increased the convergence time of full non-coplanar
   calibration.

   Bug reported by Torfi Thorhallsson <[e-mail]>


----------------------------------------------------------------------------

Version 2.0  17-Feb-94

1. Fix bug in routines ncc_compute_R and ncc_compute_better_R.
   The r4, r5, and r6 terms should not have been divided by cp.sx.

   Bug reported by Volker Rodehorst <[e-mail]>

2. Modified undistorted_to_distorted_sensor_coord () to do the conversion
   by algebraically solving the cubic polynomial

	Ru = Rd * (1 + kappa1 * Rd**2)

   using the Cardan method.

   Code donated by Frederic Devernay <[e-mail]>.

3) Added two new routines for calculating calibration error statistics:
   object_space_error_stats () and normalized_calibration_error ().

   The first routine calculates the statistics for the magnitude of the
   distance of closest approach (i.e. 3D error) between points in object space
   and the line of sight formed by back projecting the measured 2D coordinates
   out through the camera model.  The second routine is based on an error
   measure proposed by Juyang Weng in IEEE PAMI, October 1992.

4) Added checks for coordinate handedness problems in the calibration data.


----------------------------------------------------------------------------

Version 1.0  14-Jul-93

   First public release of code placed in the Vision List Archive SHAREWARE
   subdirectory on FTP.TELEOS.COM by Jon Owen <[e-mail]>.


----------------------------------------------------------------------------

Version 0.X  30-Nov-92

   First version of code written at CMU.