camera_calibration: Improve YAML formatting, make config dumping methods static

[valgur]

Some more minor improvements:

• Added from __future__ import print_function to make sure the printed output looks as expected, rather than e.g. ('D = ', [0.8272, ...]) on Python 2.

• Improved matrix formatting in the YAML output by making use of array formatting from Numpy.
• Reduced precision for camera and projection matrices to 3 in the YAML output since everything
  past 1 decimal place is mostly excessive.
• Made calibration config dumping methods static to allow them to be used conveniently from elsewhere as well.

What the YAML output looks like after these changes:

image_width: 480
image_height: 640
camera_name: narrow_stereo/left
camera_matrix:
  rows: 3
  cols: 3
  data: [502.992,   0.   , 250.226,
           0.   , 503.221, 245.983,
           0.   ,   0.   ,   1.   ]
distortion_model: rational_polynomial
distortion_coefficients:
  rows: 1
  cols: 8
  data: [6.798071, 140.213354, -0.003831, -0.002863, 1485.092992, 6.747936, 141.145419, 1475.894951]
rectification_matrix:
  rows: 3
  cols: 3
  data: [1., 0., 0.,
         0., 1., 0.,
         0., 0., 1.]
projection_matrix:
  rows: 3
  cols: 4
  data: [507.04 ,   0.   , 248.912,   0.   ,
           0.   , 505.216, 243.459,   0.   ,
           0.   ,   0.   ,   1.   ,   0.   ]

[SteveMacenski]

Reduced precision for camera and projection matrices to 3

I disagree with that, this may be used in really high resolution camera situations were there may be more meaningful sig figs to look at. If we have some numbers, I'm not sure its up to the printing area to truncate them. While I'd generally agree there may be some value to cut off at 3 actually seems too few. I'd think ~5-6 being the minimum before truncating

[valgur]

Ok, I can undo the camera/projection matrix rounding. Just to be clear, I assume you meant 5-6 figures after the decimal point rather than significant figures?

[valgur]

Set the number of decimal digits for intrinsic and projection matrices to 5.

image_width: 480
image_height: 640
camera_name: narrow_stereo/left
camera_matrix:
  rows: 3
  cols: 3
  data: [502.99207,   0.     , 250.22621,
           0.     , 503.22066, 245.98286,
           0.     ,   0.     ,   1.     ]
distortion_model: rational_polynomial
distortion_coefficients:
  rows: 1
  cols: 8
  data: [6.798071, 140.213354, -0.003831, -0.002863, 1485.092992, 6.747936, 141.145419, 1475.894951]
rectification_matrix:
  rows: 3
  cols: 3
  data: [1., 0., 0.,
         0., 1., 0.,
         0., 0., 1.]
projection_matrix:
  rows: 3
  cols: 4
  data: [507.04044,   0.     , 248.912  ,   0.     ,
           0.     , 505.21561, 243.45872,   0.     ,
           0.     ,   0.     ,   1.     ,   0.     ]

[SteveMacenski]

@valgur looks good, yeah you got my meaning sorry for the late response, I've been caught up in another repo.

This all looks good to me now, have you tried loading this output yaml and make sure it works for the other application?

Then I'll merge

[valgur]

I tested loading the YAML output as a CameraInfo message with camera_info_manager and camera_info_manager_py. Both return correct results.
Besides that, I have also been using similarly-formatted calibration files in deployed code without any ill effects.

[SteveMacenski]

Sounds good!