[GRASS-user] rectifying digital images

sebastian sauer scb at lo-res.org
Sat Jan 5 15:31:57 EST 2008


Sat 05 Jan 2008 17:13, sebastian sauer wrote:
> Sat 05 Jan 2008 16:45, Vincent BAIN wrote:
> > If you follow Markus' how-to, you may be interested in technical info
> > concerning your camera :
> > 
> > Kodak DCS Pro SLR/n/c
> > imager size : 36.0mm x 24.0mm
> > pixel size : 7.9µm x 7.9µm
> 
> exactly. even if you (for unknown reasons) do not follow Markus' how-to,
> you will need this information. and it won't hurt double-checking the
> correctness of this data against, the DPI in the EXIF data.
> 
> given that, it's pretty straight-forward to set-up precise fiducial
> marks.

turns out this is essentially known to the experts :)
cf. http://www.elcovision.com/e_kamera.html
"[DSLRs] have very low deformations; a réseau grid plate for compensation 
of film or paper deformations is not necessary. For the definition of a photo
coordinate system the four photo corners are sufficient. This means:
Digital camera do not need to be modified, but can be calibrated and used
for measurement as they are."

nb.: the markings on a réseau grid plate [0] are the fiducial marks (aka réseau 
marks.)

[0] http://www.elcovision.com/img/Reseauplatte.jpg

cf. http://www.ee.oulu.fi/~jth/calibr/ 
"Modern CCD cameras are usually capable of a spatial accuracy greater than
1/50 of the pixel size."

> 
> (cf. http://edclxs22.cr.usgs.gov/osl/figure2.gif && 
> http://grass.itc.it/grass63/manuals/html63_user/photo.camera.html )
> 
> *but* what is tricky is getting the calibrated focal length right.
> (and this of course (distortions!) is a quite important parameter)
mmh, i'm still skeptical, as i frankly really have no glue how accurate
the EXIF data really is. (hints are highly appreciated! e.g. would be
interesting if this is within the accuracy of the autofocus error of the
lens. )

but according to the experts at least decent DSLR cameras seem to produce
a quite accurate EXIF focal length data.
cf.: http://www.elcovision.com/e_elco_reseau.html

BTW, does anyone know how the EXIF focal length is defined?

> > Le samedi 05 janvier 2008 à 23:25 +0800, Maning Sambale a écrit :
> > > Camera:  Kodak Pro DCS 14N digital SLR camera 
> > > using a 19-35mm Tamron zoom lens set at 19mm focal length
> ok, his info is not so shabby, at least you know the "calibrated focal
> length" should be ~19mm.
> 
> also the spec of the lens should include the focal length tolerance.
seems which data is hard to get for most (D)SLR lenses.

anyway to get a "feeling" what is roughly possible for a single lens.
Focal Length Tolerance: ± 0.5 %
http://www.rainbowoptics.com/Spherical_Lenses.aspx

but of course a (D)SLR lenses is a cascade of many lenses. in
professional analgue photogammetry, Focal Length Tolerance: ± 3 mm @153mm 
(so this should be ~ 4.6 % )
http://erosproject.cr.usgs.gov/osl/reqtable.html 

BTW, does anyone know what camera-sizes are usually used in aerial
photogammetry? aka how big is the plate? medium-size? large-size?

> ps: if anyone knows a good trick to do a precise focal length
> calibration, plz don't keep it as your secret, but share it with us :))
> SCNR.
http://www.google.com/search?hl=en&q=camera+calibration&btnG=Search
http://www.stoske.de/digicam/Artikel/verzeichnung.html
http://wiki.panotools.org/Lens_correction_model


More information about the grass-user mailing list