[Elphel-support] Camera Calibration

[Andrey Filippov]

On Fri, Jul 13, 2012 at 6:05 AM, Sinaya Dayan <sdayan at njrati.org> wrote:

> Hello Andrey,****
>
> ** **
>
> I got a response back from our calibration people who suggested that
> Elphel preforms the calibration and provides us with the .CAL file with
> some minimum requirements. Like yourself, they do not know why the
> calibration is not successful, they have been calibrating many cameras but
> this is the first time they are unable to complete the task. At this point
> we need solutions since we cannot proceed with our project as planned. We
> are considering replacing the camera if we are unable to perform the
> calibration but would really like to explore all options before making a
> decision. We intend to use it for aerial image collection and the software
> we use for processing requires calibration. Without it nothing can be
> accomplished.
>

Hello Sinaya,

I'm very sorry you are not able to use Elphel camera because you were not
able to calibrate it. I tried to explain how I see the reason - the lens
you use is a high quality (and low distortion) one and it is oversized
(2/3" format for the 1/2.5" sensor) for the sensor so it really is a
challenge to measure distortions, and the targets you tried to use for it
are not adequate. It is also understandable that people who processed your
files could not understand the problem - normal cameras do not have
oversized lenses so they all have visible distortions especially in the
very corners of the image, but the corners of 1/2.5" sensor correspond to
the nearly center region of the 2/3" lens.



> ****
>
> Would you be able to perform the calibration for us and send us the file?
> (I will forward a list of our requirement s as soon as MosaicMill sends it
> to me).
>

No, we do not offer such service for all lenses, currenlty we only
calibrate f=4.5mm lenses we use for Eyesis. And we use different
calibration method and calibration target (we need higher precision) so we
use different format and can not provide you with .CAL files.


> ****
>
> Is there another way to communicate with you other than the support list?
>

Sales matters are handled through our ticket system and phone, but the only
technical support we provide is over the mailing list.

Andrey


> ****
>
> Looking forward to hearing from you and thank you much for all your help,*
> ***
>
> Sinaya****
>
> ** **
>
> *From:* elphel at gmail.com [mailto:elphel at gmail.com] *On Behalf Of *Andrey
> Filippov
> *Sent:* Monday, July 09, 2012 1:14 PM
>
> *To:* Sinaya Dayan; support-list at support.elphel.com
> *Subject:* Re: [Elphel-support] Camera Calibration****
>
> ** **
>
> Hi Sinaya,
>
>
> There is nothing to move inside the camera (if the lens is attached
> firmly). I'm not sure - what kind of processing do they make - is it open
> source or proprietary program? Do the results depend on the quality of the
> target, sharpness of the images and flatness of the target (assuming there
> is no wind so the target shape does not change between the shots)?
> Can the people who perform calibration provide you with some numbers -
> what are acceptable deviations of the target spots from their ideal
> position? What is the acceptable non-flatness?
> My guess is that the lens is high-quality, low distortion one (and it is
> 2/3", not 1/2.5" so the distortions inside the smaller sensor area are even
> smaller than specified for the lens). Measuring distortion parameters of
> the low distortion lens is much harder than measuring parameters of the
> lens with the higher distortions. If you had an ideal no-distortion lens
> you will not be able to find even the position of the lens center!
> So I believe that the distortions of the lens are lower than the quality
> (geometry, flatness) of the target you use (the focusing quality also
> varies a lot) and that leads to erratic results. We have our target printed
> on much more solid material and we still have to calibrate the target
> itself (it is 3.022m x2.667m), because with the f=4.5mm lens we can detect
> ~0.1mm shifts of the pattern, and the printer and material stretching can
> not provide such precision.
> So I would recommend you to print the target on a really wide paper (at
> least) and make sure it is flat - do not use ink jet printer that made your
> target non-flat with large waves. Than you may be able to measure
> distortion, but I'm not 100% sure that this method will work without target
> calibration.
>
> Andrey****
>
> On Mon, Jul 9, 2012 at 5:51 AM, Sinaya Dayan <sdayan at njrati.org> wrote:***
> *
>
> Hi Andrey,****
>
> Here is one of the image sets we took along with the target measurements
> and its calibration report.****
>
> The analyst approved that the images are good for calibration. We have set
> the focus to infinity (@48.5 mm) the entire time and the target size was 3
> x 2 meters. We have not changed anything in between images part from the
> camera’s position. ****
>
> “Hardware problems” according to the analyst could be a small internal
> movement that could jeopardize the image and result in an in consistent
> error. ****
>
> Please have a look and let me know what you think.****
>
> Thanks,****
>
> Sinaya****
>
> ** **
>
> _______________________________________________
> Support-list mailing list
> Support-list at support.elphel.com
> http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20120713/319ca021/attachment-0002.html>