Augmented Reality picture detection

Hi evvvverybody,

I begin a project with augmented reality.
I would like to use vvvv because I want to link my project with diferent stuff and “I’m use to” play with this nice software.
But I dont want to use fiducial or ARToolkit tracker for this project. I want to know if it’s possible to recognise an other shape quite ?(More complexe than just an ugly black square).
Ideally I would like to recognise a picture but it’s probably very difficult to do that…

Thank’s by advance for your advices,

Roman

hi roman, you need to have a look to opencv Haartraining possibilities.
dont know how to use it inside vvvv

yes,that’s quite difficult if you want recognise any pictures as mark.
maybe you can find some more info from here http://cvlab.epfl.ch/software/index.php

hi roman,
what you are looking for is called natural feature tracking, but unfortunately for now there’s not way to do it in vvvv (or i don’t found ti for now).
take a look of this link to understand better how it works: http://studierstube.icg.tu-graz.ac.at/handheld_ar/nft.php

you could cover ARTK-markers with some filter for visible light so that they appear black.

then use IR-light and IR-cam for tracking

i think i even heard about paint that is transmissive for IR only.

some IR ink links,

http://www.nightmarker.com/products/ir-paint
http://www.maxmax.com/aXRayIRInks.asp

was dreaming about IR paint ;]

@lasal:
cool tip, thx!

you can also use cheaper “magic” or “reveal” marker, but only the white one, used to reveal the secret colors, wau!

Just a quick thought and a quick patch - you could check it by this approach:
Take all rotations and all sizes (?) of you pic and adding it with your liveimage. If it matchs at a certain value you can get its rotation an actually know that its your pic.
If so, best with shader as it does multple pics at on time.

Anyway the keypoint of this is to extract your desired target area irst - that is what the LFO rotating pic would be.

After overthink it, I guess its too unprecise… But may be others may like it.

RotationPic.zip (12.5 kB)