
----------------------------------------------
DLR CALLAB -- HOW-TO MANUAL -- BRIEF TUTORIAL
----------------------------------------------


   1. Preparation.

      Run DLR  CalDe  or  any  other  similar  program  to detect  and
      identify features of a particular calibration object.

   2. Load data.

      Load the PNG  image files as  in  DLR CalDe. The  application is
      flexible to  names and numbers used.  At  least three images (or
      stereo  image pairs) are required.  You may select more than one
      file at  the same time by  pressing the  SHIFT key and selecting
      the first  and the last file,  or by pressing  the CTRL  key and
      selecting each file. Load the corresponding points files for the
      images as from DLR CalDe.

   3. 1st Calibration Stage.

      The  general  settings  should  not be  modified  in a  standard
      calibration  case.   If  you are   ready for  camera  parameters
      estimation (take  a look at the  command line output) do proceed
      with the 1st   calibration stage.  Intrinsic  camera  parameters
      (including  the camera-#1 to  camera-#n transformation(s) in the
      case  of a n-stereo  camera rig) are about  to be estimated.  At
      the   beginning    a first estimation     takes   place based on
      homographies (refer to the work of  Zhang or Sturm and Maybank).
      Then,  the estimation is  numerically  optimized. After that the
      point projection  errors are overlaid   on the images.  You  may
      browse    through the images in order    to  look for mistakenly
      detected points.  For   this, it is  useful to  hide the  actual
      image  (click on 'switch monochrome  image')  and to augment the
      error arrows (click on 'error zoom').  In  addition, you may use
      the histograms on the right-hand  side for rapidly finding wrong
      points.  Repeat this 1st  calibration  stage if you  happened to
      remove any point.

   4. 2nd Calibration Stage.

      The  2nd    calibration   stage  supplies  the      hand-eye, or
      TCP-to-camera transformation.  This transformation is not always
      required by  and can only be  estimated if the TCP-to-robot-base
      transformations are available for every calibration image. These
      transformations  were  processed by DLR  CalDe  and ought  to be
      embedded  in the  points    files.  DLR CalLab  offers   several
      algorithms          for     estimating  this      transformation
      ("Settings->General settings"):

        a) On the   one hand you  may use  the  method  of  Strobl and
        Hirzinger  [Strobl06]  minimizing  errors   in  the  erroneous
        world-to-TCP  transformation.   The algorithm  uses  a  linear
        least squares solution for the first estimation.

        b) On the other hand  you may minimize the features projection
        errors.  This is only a  sound solution in  the case of highly
        noisy  cameras  with very  low    image resolution, and   very
        accurate world-to-TCP transformations.

   5. 3rd Calibration Stage.

      The 3rd calibration stage verifies the correct estimation in the
      last steps and furthermore produces accuracy estimations for the
      positioning device (e.g.  robotic manipulator, infrared tracking
      system, etc.)  responsible for the world-to-TCP estimations.

   6. Save data.

      "File->Save" saves the calibration results in the desired output
      format.  For more information  refer to "Help->Documentation" in
      DLR CalLab.



















































