next topic »

[on:]

Hi All,

For a future project I'm looking for a way to have a mobile(and preferably small) face detection system.
I'm wondering if there is a way of doing face detection with a stand alone arduino?

Or, if there should be computer involved doing the calculations, does anyone know about a good, reasonably priced wireless camera(also as small as possible).

I've been searching a lot but couldn't find anything usefull.
Thanks in advance,
Rob.

[Reply #1 on:]

It should be possible to arrange a set of microswitches into different depths so that if a person presses their face against the differently depthed arrangement it would trigger the microswitches all together. For example, microswitches arranged where eyebrows, brow, cheeks, chin, nose are expected to be in their respective position and depth. If you press a part of anatomy, or a non-face object such as a carrot, against the arranged switches, it can deduce that it isn’t a face. If all the switches trip, then it had detected a face.

[Reply #2 on:]

Using a camera I'm afraid it won't be possible with an arduino, it can do a lot, but lacks the speed necessary and has too little memory. A PC or ARM-board will be better for such a job.

I have little experience with them, but guess wireless webcams exist for surveilance etc.

[Reply #3 on:]

For the wireless camera, you could consider trying this. It appears to be small, too. I have heared of people using processing for image recognition, so you might be able to do something with that.

Onions.

[Reply #4 on:]

Have a look at this
http://sophiateam.undrgnd.free.fr/microcontroller/camera/index.html
The gameboy camera has built in edge detection amongst other things, same method used to do OCR and read vehicle number plates, speed need not be an issue, the camera module uses an SPI interface all you have to do is set the pixel clock and read in the analog value for that pixel

[Reply #5 on:]

Hi!

Thanks for you replies.
In the meantime I did some research. And this comes very close.
http://cmucam.org/

Rob.