I downloaded a fairly newish OpenCV Processing and Java library found at :
http://www.ubaa.net/shared/processing/opencv/
and looked at the (less than 100 LOC) blobs example using my development FTIR mini-table. I have played with this for less than 5 minutes but wanted to share because I think this is pretty cool. If you take a look at the screenshot I took, you can see that my table is pretty “dirty” looking with an old and bubbly layer of SORTA-Clear but blob detection was pretty resilient. I should have played around with the threshold some more but again, too excited and eager to share with y’all.
The top two images are the same raw output from the camera. The bottom left image is used for background subtraction and the bottom right is the money shot with blob detection. Pressing down on the acrylic causes the camera to see the FTIR from my bubbly SORTA-Clear (I guess my poor hardware setup shows the resiliency of the blob detection algorithm).
OT : Also check out the face detection sample, it worked pretty well for me out of the box. I am thinking that if a camera were positioned correctly, with the aid of a face detection algorithm, you could effectively track eye movement. Shining IR at someone’s face, the pupils become like blobs that can be tracked? It would be cool to do eye tracking . . . but I digress, too many multi-touch app. ideas floating in my head . . . need to start integrating this with the Processing TUIO library or straight to Flash (http://processing.org/discourse/yabb_beta/YaBB.cgi?board=Integrate;action=display;num=1115427973). Again, need to some leg work before coding.
Anyway, hopefully someone finds this useful.
