CameraCalibration Export parameters

[Paulo]

Dear Support,

I am noticing that when I export a Camera Calibration, the sign of radial and decentering distorsion parameters are inversed when I chose Inpho versus Australis format.

In the example I have Camera0_SN102146 whose PhotoScan Adjusted parameters are given in attached screen copy.

Now if I save the Camera parameters as Inpho calibration, I get following file:

$CAMERA
  $TYPE : Camera0_SN102146
  $DATE : 16:33:36 30/01/2016
  $BRAND : Custom
  $KIND : CCDFrame
  $CCD_INTERIOR_ORIENTATION :
     135.135       -0    2449.09
       0.0000000000    -135.135    1642.23
  $CCD_COLUMNS : 4872
  $CCD_ROWS : 3248
  $PIXEL_REFERENCE : TopLeftTopLeft
  $FOCAL_LENGTH :    65.0375
  $PRINCIPAL_POINT_PPA :     0.000000     0.000000
  $DISTORTION_TYPE :  Polynomial
  $RADIAL_COEFFS :
                      0            -7.22631e-006                  1.5452e-008                  -4.18738e-012
                      -6.84534e-017                  0                  0                  0
  $DECENTRE_COEFFS :
                     3.45261e-006                  -6.56677e-006                 0                  0
  $GPS_ANTENNA_OFFSET :     0.000000     0.000000     0.000000
  $CAMERA_MOUNT_ROTATION :     0.000000

and If I save camera parameters in Australis format, I get:

Adjustment: Free-Network
Number of Points: 1000
Number of Images: 10
RMS of Image coords:    1.00 (um)

Results for Camera 1    Camera0_SN102146      Lens

Sensor Size        Pixel Size (mm)
  H    4872           0.0074
  V    3248           0.0074

  Camera    Initial      Total          Final        Initial         Final
 Variable    Value     Adjustment       Value       Std. Error     Std. Error
    C       65.0375      0.00000       65.0375       1.0e+003     1.000e-003 (mm)
   XP        0.0968      0.00000        0.0968       1.0e+003     1.000e-003 (mm)
   YP       -0.1349      0.00000       -0.1349       1.0e+003     1.000e-003 (mm)
   K1  7.22873e-006   0.000e-000  7.22873e-006       1.0e+003     1.000e-003
   K2 -1.54725e-008   0.000e-000 -1.54725e-008       1.0e+003     1.000e-003
   K3  4.25211e-012   0.000e-000  4.25211e-012       1.0e+003     1.000e-003
   P1 -3.45260e-006   0.000e-000 -3.45260e-006       1.0e+003     1.000e-003
   P2  6.56677e-006   0.000e-000  6.56677e-006       1.0e+003     1.000e-003
   B1 -6.91731e-008   0.000e-000 -6.91731e-008       1.0e+003     1.000e-003
   B2 -6.69269e-008   0.000e-000 -6.69269e-008       1.0e+003     1.000e-003

Maximum Observational Radial Distance Encountered:      1.000 mm

As you can see, the radial parameters are basically the same but inversed in sign between Inpho and Australis format as well as decentering parameters (green).

Could you explain this, please?

There is also a new format fot cmaera calibration, USGS. In this case it produces a file with additional K0 parameter but with Focal set to 50 while real focal is around 65 mm. Why is this so?

here is USGS format calibration saved from same camera:

Camera: Camera0_SN102146
Date: 15:37:41 30/01/2016
Width: 4872
Height: 3248
PW: 0.0074
PH: 0.0074
F: 50
XP: 0.0968388
YP: -0.134938
K0: -0.231213
K1: 5.55418e-006
K2: -1.18667e-008
K3: 3.17056e-012
K4: 1.15687e-016
P1: -2.65083e-006
P2: 5.04159e-006
P3: 2.82698e-005
P4: -7.25204e-008

Thanks for any input on this...

[Alexey Pasumansky]

Hello pap1956,

According to our information the camera distortion models in Australis and Inpho are different and the difference is caused by the inverted distortion coefficient signs. We also have the confirmed that experimentally, so if you have any project where any of these exports is not working properly, please provide the additional information.

As for the focal length value in the USGS export, it seems to be an issue on our side and we'll try to fix it in the next update.

[RLF123]

Hi Alexey/All,

I apologize up front if this has been answered elsewhere, but can you point me to the algorithm/papers that describe the conversion  between the australis calibration coefficients and the defaults shown in the Photoscan GUI?

For the radial terms, It appears that the conversion is roughly (but certainly not exact):

k1_photoscan = f^2 * k1_australis
k2_photoscan = f^4 * k2_australis
...

where f is the focal length.

Any insight greatly appreciated.

Thanks!