Lens calibrate parameters

17 views
Skip to first unread message

T. Modes

unread,
Oct 1, 2009, 6:47:28 AM10/1/09
to hugin and other free panoramic software
I tried to use calibrate_lens. I got the follow output:

<snip>
Camera digicam h 4000 (250.0/mm) v 3000 (250.0/mm)
Lens rectilinear FL 14.0 mm (3500.0 pixels)
Iter: 0, estimate: 0 -- errors 0.123578032 0.00279322129
Iter: 10, estimate: -0.0126571199 -- errors 0.00120297859
0.00123479922
Iter: 20, estimate: -0.0125559089 -- errors 6.87047212e-005
0.00123468594
Iter: 30, estimate: -0.0125510906 -- errors 3.7203293e-007
0.00123468566
Iter: 40, estimate: -0.0125510906 -- errors 3.31792542e-005
0.00123468566

Levenberg-Marquardt returned in 45 iterations
reason 2 - stopped by small Dp
start ||e||_2 : 0.00279322
final ||e||_2 : 0.00123469
||J^T e||_inf : 3.72358e-007
||Dp||_2 : 8.32767e-041
mu/max[J^T...]: 2.25079e+014
iterations : 45
stop reason : 2
function evals: 51

Fitted parameters...
Radial mean sq. dev. 0.0015 factors:
0.0000 0.0000 0.0000 -0.0040 0.0000 0.0000 0.0000

What does this 7 values mean? How does they compare to the panotools
model (how do they translate to a/b/c paramters in hugin)? Is there
some documentation?

Thomas

Pit Suetterlin

unread,
Oct 4, 2009, 3:24:58 PM10/4/09
to hugi...@googlegroups.com

Hi Thomas,

playing around with it myself (I have a new camera ;^>), I didn't get
much further than you. I'm using images of a weight-tensioned rope
taken at different camera rotation and feed them to calibrate_lens.
First I did not get any results, it needs al least 12 located lines.
Then the same as you:

Levenberg-Marquardt returned in 53 iterations
reason 1 - stopped by small gradient J^T e
.....
Fitted parameters...
Radial mean sq. dev. 0.0018 factors:
0.0000 0.0000 0.0000 -0.0048 0.0000 0.0000 0.0000

Whatever the numbers mean, they don't have much to do with lens
correction parameters, which for this lens should be

a="0.00455538" b="-0.0236561" c="0.0143293"

determined by marking the rope as (separate) lines in hugin and then
optimizing only a,b,c.

Locating the lines works fine, just looks like the LM iteration is not
succeeding :-(

Cheers,

Pit
--
Dr. Peter "Pit" Suetterlin http://www.astro.su.se/~pit
Institute for Solar Physics
Tel.: +34 922 405 590 (Spain) P.Suet...@royac.iac.es
+46 8 5537 8507 (Sweden) Peter.Su...@astro.su.se
Reply all
Reply to author
Forward
0 new messages