Progress!

on Friday, June 11, 2010

So I ditched the "table" idea for now because I don't really have the capital to build one from scratch and the one we have in the office doesn't support fiducial recognition! I've decided to improvise and do up one that utilises the top projection system that we have..


I have been sawing and drilling aluminium rods and L-brackets to make these metal cubes! I will then mount the fiducial markers (super big ones!) on a cardboard and mount on the cubes (like a dice!).

Hopefully this will work.. I am slowly getting the gist of building my own flash app.. but the linking of touchlib and reactivision is still uncompleted! -.-"

Skype-ing with the China guy. . .

on Friday, June 4, 2010

Most of the time, I don't understand what he's saying.. But I think he understands what I'm trying to say with the Endlighten acrylic and the IR light strips going around it.. He also understood what I meant with the fiducial thingy..


So what I gathered is..
1) He needs more info (specs) on the Endlighten acrylic and he can help me to find it in China where it MIGHT be cheaper cause like I said, it is fucking expensive to get it from the US/Singapore (imagine SGD1.2k for a small piece FTS)
2) He advised me to use multiple cameras (2 or more) for more accurate fiducial tracking. Using only 1 is more for finger tracking zz
3) OpenCV for camera capture (I know nuts about this)
4) OpenGL for 3D graphics for MT apps (I know nuts about this too FML)

So for now, I guess what I'll be doing is to get the Endlighten, LED strips, a webcam with band pass filter, some cardboard to construct a box, and use touchlib to test it out first!

When that's working, I'll try and somehow make reactivision work on the system. Have to ask ze boss for funding. zzzz


Development of a Multi Touch System with Fiducial Recognition

I'll be using this blog to document the process of this whole project's development.. Basically I have to get this running before running back (off) to Singapore in August.. (At least get the hardware and some simple software running first..)


So I've decided for now to use the Diffused Surface Illumination (DSI) technique for my future (non-existent as of yet) MT table. Why DSI? Because it is only 1 of 2 techniques that makes fiducial (or any other objects for that matter) tracking possible. The DSI setup is much simpler too compared to the other one, but it is nearly impossible to get Endlighten acrylic here (and it's fucking expensive too).

I have drawn up a list of materials that I will need for the hardware:
1) Projector - Use existing one in the office?
2) Camera - Use a cheap webcam? Will have to dismantle and remove the IR filter (or use sandpaper to file it away =/)
3) Band pass filter lens for the camera (850nm? depends on the wavelength of the IR LED strips that I can get)
4) IR LED strips (850nm? seem like it works better than the 900+nm ones..)
5) Endlighten acrylic/plexiglass - Ship from the US? Or Singapore? or just use regular acrylic and get some diffusing compliant material (like silicone, which is a bitch to layer on perfectly on the acrylic) sandwiched between the acrylic and another thinner layer.
6) A nice casing to house it all in! I haven't decided on the layout of the entire setup, but depending on the situation (presence of hot spots/ghosting), I might have to get a hot mirror(to eliminate hotspots) and/or a front surface mirror (reduces the distance for the projector and camera and also reduces ghosting)

For the software.. it seems that touchlib is the way to go.. but I need something that can do fiducial tracking AND finger tracking.. reactivision can do the fiducial part, but it sucks in terms of finger tracking (and is harder to do the interfacing). Ah, the DILEMMA! And I'm getting so frustrated trying to code in Flash and ActionScript! I'm a newbie in this field unfortunately =(

FML, I thought the programming nightmare would've ended when my Major Project was done.. ='(

References: