3.5 update and projector calibration error

[Rogier van Stapele]

Hi,

I updated to version 3.5. Newer is better

In previous versions I could do the camera and projector calibration from a set of 27 images (1 for the camera, 26 for the projector). Now the projector calibration results in an error.

As far as I can see the projection images are still the same, so are the settings. What happened?

Rogier.

[Sven]

Hi Rogier,

Whoops! Please go to Settings -> Advanced Settings -> Structured Light and set NoiseReduction to 0.
NoiseReduction does not work with ImageGrabber.

Sorry about this - you could not have known. We'll think of a solution or at least a helpful error message.
Edit 2012-10-18: Problem fixed for next version

Please let us know if this solves your problem.

Sven

[Rogier van Stapele]

Hi Sven,

The error has changed now:

ImageSequencing::OnTimer(): Sequence recording finished
delta t statistics: num=26, min=0.159811, max=0.244635, avrg=0.195996, deviation
=0.0249461
Grabbed 27 images in 5.09589s --> 5.10215 FPS
~
~
~
Size of camera's center pixel at background plane: 0.512432 mm
cy=-0.5 rmsError=0.0160456
cy=0 rmsError=0.0160455
cy=0.5 rmsError=4.05376
cy=1 rmsError=0.0182108

Warning: undistorted image point to distorted image point mapping limited by
maximum barrel distortion radius of5.80504
(xu = -0.00498308, yu = -3.87022) -> (xd = -0.00747424, Yd = -5.80504 )


Warning: undistorted image point to distorted image point mapping limited by
maximum barrel distortion radius of5.80504
(xu = -0.00146967, yu = -3.87201) -> (xd = -0.00220337, Yd = -5.80504 )


Warning: undistorted image point to distorted image point mapping limited by
maximum barrel distortion radius of5.80504
(xu = -0.00262428, yu = -3.87151) -> (xd = -0.00393492, Yd = -5.80504 )


Warning: undistorted image point to distorted image point mapping limited by
maximum barrel distortion radius of5.80504
(xu = -0.000690132, yu = -3.87073) -> (xd = -0.00103501, Yd = -5.80504 )


cy=1.5 rmsError=0.0365767
9249 outliers removed (total 10158)
~
Projector calibration failed!

- It reports to have read 27 images where I selected only 26.
- I enhanced the white balance of all images. The lines show as perfect sine/square patterns, without clipping in black or white.

Does this help?

Rogier.

[Sven]

- It reports to have read 27 images where I selected only 26.

That's normal, don't worry about it. Version 3.5.1 will always show "1 too much".

- I enhanced the white balance of all images.

You should not apply any filters to the images... Did scanning work in DAVID 3.4 with these images?

When scanning with ImageGrabber, you must have the same settings as during projection. Obviously the "Number of patterns" that DAVID shows below the Start button must be identical. Also the values for Frequecies, CodedLight-Inverse, PhaseShift-Shifts and -Multiplier must be identical.
Usually they will be the same. However the "auto" values can be different when you have a different projector resolution or camera resolution setting.

I have a set of SL pattern images which I use with the ImageGrabber. They were made with 3.4 but also work fine in 3.5.

Sven

[Rogier van Stapele]

Hi Sven,

I reused the original scan images (the ones that previously worked) with this error as result:

Grabbed 27 images in 3.82241s --> 6.802 FPS
~
~
~
Size of camera's center pixel at background plane: 0.51243 mm
cy=-0.5 rmsError=0.011374
cy=0 rmsError=0.0113738
cy=0.5 rmsError=0.0113737
cy=1 rmsError=0.0121053
cy=1.5 rmsError=0.0121056
8556 outliers removed (total 10239)
~
Projector calibration failed!

I checked that the used projector images are identical to the new ones: 7 hor. lines + 6 hor. shifts + 7 vert. lines + 6 vert. shifts = 26 in total.

I noticed that the setting StructuredLight/Frequencies=8 results in 7 patterns. Why?

Do you have a 3.4 version that I can use to recreate my setup?

Rogier.

[WalterMo]

Regarding DAVID 3.4.0:

See Sven's first post of Oct 17th:
viewtopic.php?f=15&t=5330&p=24791#p24791

Walter

[Rogier van Stapele]

Thanks Walter.

I tried it with 3.4 and the calibration works, in 3.5 it doesn't.

Here are the settings I use:

Orientation: both
mode: cl+ps
Autofreq.: yes
Frequencies: 8
Brightness: 255
AutoPatternHoldingtime: yes
PatternHoldingTime: 250
MinContrast: 6
MinTrianulationAngle: 15
CodedLight/Inverse: No
PhaseShift/Shifts: 6
PhaseShift/Multiplier: 3

[Sven]

Hi Rogier,
If you post a link to your SL image sequence, I can test them with the ImageGrabber and try to find out why it does not work with 3.5.
Sven

[Rogier van Stapele]

Hi Sven,

Posting the images didn't work because the zipfile is larger than 2 MB. I'll send them by mail.

I put both the projection images and the captioned images in the zipfile.

Rogier

[Sven]

Hi Rogier,

I have tested your images. I am not sure why your images work in 3.4 and not in 3.5.

However the problem is that your image shows "too much" around the calibration corner. The camera should only see the calibration panels (only the regions where there are markers). You should zoom in, or move camera (and projector) closer to the scene, or use a larger calib-corner. This depends on the size of your object, which should (nearly) fill the camera view and the projection size.

I am sending you new images by mail. They are your images, but I have cut away the borders then scaled back to the same size. Poor quality, just as an example.

Sven

[Rogier van Stapele]

Hi Sven,

You are right, there was too much space around the pattern. I assumed that only the space within the square raster mattered and apparantly the previous versions where less critical.

This new knowledge confronts me with a new physical problem because I am scanning very close to the floor, but I'll find a solution for that.

Thanks very much for your help. Btw the new function to save all pattern images at once is very nice, it saves me an hour of pressing print screen and saving.

Rogier.