hallo guys,
i definitely need some help. I try to install an multitouch for an holographic screen, which is nearly just an glass pane without special optical attributes. So my problem is that i have no dispersive/diffusing element in my construction. The object i want to install the multitouch to is an glasern lectern. It has two angels so total reflection would not work with Lasers on the bottom, i think. because of the great looking glasern optic i would like to avoid any cables to be around the lectern.
I already installed a common Di setup. But i was not able to detect blobs all over the screen, only in the area I made the touchlib adjustments. A big problem is that the hole hand is in the cameras picture and i am not constantly able to select fingertips out of it.
Hopefully someone has experience with similar problems, or anyone has a good idea or tipp.
I’m thankful for ever answer!

This is how it look like:

Perhaps DI isnt the best solution for your application.  I would suggest either FTIR or laser techniques which would solve your problem.

Yea LLP (Laser Light Plan) would be your best bet to start....

Try starting here: http://nuigroup.com/forums/viewthread/2178/

Or searching forums for LLP

hey holo,

found the source i have mentioned in the PM: http://www.roborealm.com/downloads/version_log.php
Maybe that is a starting point/inspiration for your specific problem. Since you want to have your construction completly transparent and you don’t want to sacrifice your design, I think that your only oportunity is to track the whole hand just like the guys at multitouch.fi are doing.

Cheers,

I would think LLP would work since it doesn’t require a diffuser or compliant surface.

The guys at multitouch.fi are doing DI. Since with good lighting in DI it makes it possible to see the whole hand and be able to track it. Stereovision (2 cameras) would also allow you to track the whole hand. Microsoft’s ‘touchlight’ uses that technique with a transparent screen. You might want to look up Microsoft/Eon Reality Touchlight to see a video.

LLP would be the easiest out of those methods.

cerupcat: yes sure they are doing DI. But since there is no diffusor - no DI But I just was talking about the hand tracking method (edge detection or even stereo). This seems to be for me one of the methods… Regarding LLP: sorry, but i have some problems with the usage of “home-made” IR-Lasers in public areas. I doubt also that here in germany you would get a TÜV approval for such an instalation…

But hey! Only my2c

For the scenario presented i’d go with llp its the most easiest to create not only that its the most easiest to maske in a clear display.

This is off topic, but sandor: the MT cell does have a diffusor =)

great, you guys dealt with my problem, thanks a lot !
@ Sandor: thank you for the link, i surely will try that presently.

well i was searching the forum for a while and found this in the topic “What is a diffuser? “ :

cerupcat - 16 September 2007 01:13 PM

cicassazn, are you using Diffused Illumination?

If so, then one of the problems you’re having is you have touchlib set to using FTIR instead of DI. In DI you need to edit the config.xml file (open with notepad) and take out the contrast and brightness filters. Go to touchlib, and open the config.xml file. Where it says “filtergraph” delete everything from that point on and replace it with…

<filtergraph>
    <dsvlcapture label="dsvlcapture0" />
    <mono label="mono1" />
    <smooth label="smooth2" />
  <backgroundremove label="background4">
        <threshold value="0" />
    </backgroundremove>
    <highpass label="highpass">
        <filter value="10" />
        <mode value="1" />
        <scale value="2" />
    </highpass>
    <rectify label="rectify5">
        <level value="25" />
    </rectify>
</filtergraph>

These are the filters you need for a DI setup. The way you change settings is by moving the sliders around until the rectify filter shows good results. Before you start moving sliders around, click “b” first to subtract the background from the first image, and then start changing the sliders to your desired setting.

And this really helped me. I can now detect fingertips separated from the rest of the hand. A problem I have by now is that detection works in the center of the screen better then at the outside, but I guess this can be removed by using a more powerful IR source which lightens the display more homogeneously.

cerupcat - 03 September 2008 01:00 PM

This is off topic, but sandor: the MT cell does have a diffusor =)

cerupcat: yepp you are right. I meant the installation of holo It is just my “beautyfull” english what leads to this absolutly clear statements from my side

@ sandor: haha. It’s ok =) We’re on the same page now

@ holo: I’m surprised those settings worked so well. You will definitely get a lot of false inputs this way, but it may work well enough for what you need. I think using more illuminators could help your unevenness problem.

cerupcat - 03 September 2008 01:57 PM

@ holo: I’m surprised those settings worked so well. You will definitely get a lot of false inputs this way, but it may work well enough for what you need. I think using more illuminators could help your unevenness problem.

Yepp: and that makes the difference: you can go ahead and get some clear and bright results (i mean: look ma, there are blobs on the screen!!!! Yipppiiie!).... But: and that is (only my2cents worth) experience - if you cannot achieve something like a constant visual input without the changing environmental influences (your hands, the hands of others, the ambient light, the sun and the moon, speedy gonzales and other friends moving around your setup) you will get on the long run just one thing: headaches.

Most of the people (including my genious person) are doing in the beginning the same mistake: take that cool touchlib, make some good and clear blobs by pressing “b” for “bg.removal” everytime you get some false positives and after that---- wait for the real interaction - (Geeeeeee, Maaaa.... look that, there are a LOT of blobs out there...)

I realy don’t want to dicourage here. I’m just stating that you can not remove that massive background with the currently available version of TL:

1. if you don’t have enough contrast in your picture
2. if you have some changing /moving parts in your picture (remember speedy and friends!)

Why? With the current version you will allways compare the movement of some bright pixels against a dark background.

But again: this are only my2c

Groovy!

Why not plain FTIR, without a compliant surface? Maybe it’s enough for the intended purpose.
There’s that transparent touch foil too (DISPLAX, etc.), but I believe it isn’t multi-touch, right?

edit: Never mind, I’m reading the post again and I see you don’t want cables around the surface? That would be tricky, the only thing that comes to my mind is getting into more complex computer vision stuff (like the stuff cerupcat mentioned, stereoscopic vision, etc.).