[sane-devel] How i can use 2 or more scanners at the same time??

Jeremy jeremy at acjlaw.net
Mon Jul 21 19:04:23 UTC 2008


On Sunday 20 July 2008, m. allan noah wrote:
> On 7/20/08, kid2k4 at email.it <kid2k4 at email.it> wrote:
> > Hi to all (sorry for my english)
> >  I write from Italy, with other young artist are working on a machine
> >  that requires a man-size scanner. An A1 or A0 scanner it's very
> >  exspensive for us, ahah.
> >  I use Gnu/linux (Debian-like distro) but i'm not i geek, :-( However i
> >  know the potential of open source software. I thought immediately to a
> >  solution ad-hoc but i don't know if it's possible. The idea is about a
> >  big scanner builded with more (6/8). We thought to disassemble scanners
> >  and take only the parts needed to scan.
> >  I make us some questions.
> >
> >  - It's possible to control more that one scanner with a tool like SANE?
> >  - If yes, it's possible to build a single scanned image from the scanned
> >  images of each machine,  in real time :-)
> >  - If we try to make the scanner on lan an then take each scanned image
> >  and build one big images? We need one pc for each scanner?
> >  - More and more.... :-)
> >
> >  I am asking a bit around and seemed a good idea to also ask the
> >  community of SANE.
> >  I hope the this idea can make some step to our goals.
> >  Bye bye.
>
> 1. the optical package of each scanner cannot be placed end-to-end and
> get a complete scan, because there is some dead space for packaging or
> mechanism on the ends. so now you have to stagger them, and then have
> a staggered calibration strip. correcting skew and adjusting height
> will be excercises in very precise machining.
>
> 2. you will have to have much larger and very rigid guides and carriage.
>
> 3. you will have to have a much larger motor, or convince the series
> of motors to move at EXACTLY the same rate, or some of them will be
> dead weight (very difficult to do if they have independent
> controllers)
>
> 4. you will have to multiplex all those streams of data, probably
> causing dataloss or motor stoppage/backtracking when the pc cannot
> keep up.
>
> as you can see, software is the least of your problems :)
>
> i suggest that you change directions. build a very square and sturdy
> moving platform, and scan with one scanner, using multiple passes,
> shifting the work one width left each time. you would have to stitch
> the pieces together afterwards, but that will be easy.
>
> allan
> --
> "The truth is an offense, but not a sin"

I've used hugin <http://hugin.sf.net> to stitch together  overlapping scans 
into one single larger composite image. The website has a tutorial on  howto 
stitch together the scanned pieces. It's a tedious process, since for each 
pair of overlapping images, you must select control points (effectively 
initial guesses) that the image fitting algorithm uses to optimize the 
parameters in its image fitting equation. There is an "Auto-stitch" function 
which can try to automatically find similar control points for each pair of  
images, but this has never worked for me as the algorithm finds too many  
false local minima. For production use, say in an architectural engineering 
firm, hugin would never replace an extra-wide format scanner, but it's 
adequate for the occasional one-off composite scan. And if you want  to 
create  composite montages of photos, e.g. a 360 degree view of your cityline 
or  Mars' horizon, or a hi-res photo of some planet, then hugin or some other 
image stitcher is of course necessary.

-- 

Jeremy Johnson



More information about the sane-devel mailing list