Thread: computer vision stuff

Originally Posted by KaZeR

This is really amazing.

Maybe you should try xvidcap or something like that, it would probably make better videos?

You are absolutely right, what i'm doing with my camera is to show it is real and not some vapor thing made with some video editor.

Originally Posted by shotgunefx

Interesting stuff, stop getting me thinking

As far as tracking the eyes, what about 3 (or more) IR leds around the screen, then a visible light filter over the camera, I would think you would be able to see them reflected in the iris if you were looking that way. Maybe you could use the change in the "cluster" position and size to determine eye position?

the screen surrounded with leds is a good idea but if this experiment fails then it could be possible to kill different light effects that could occur like high beam lights, sun i'm thinking now for some stuff that already exists in industry from long time, i mean an eye tracking system made of cheap hardware.

ideas :

Old glasses that can be dismantled or new one like some cheap safety glasses that can be dismantled too.
Prism, fiber optic to relay optic signal ?
Cheap pin hole camera sensitive to infra-red light or very small camera with tiny lens.
Infra-red led to enlighten the eye whole the time

Mouting a prism or a fiber optic, infra-red led for lighting, while this stuff mounted on the glasses structure to avoid the weight of the camera pcb, don't know if this would work i mean prism or fiber optic for relaying ?

If this could work then using the eye detection code added with some duration could allow a mouse move. We have the bearing, add duration, the mouse cursor could move accordingly to the time and direction with eyes or eye move.

Now to get something really usable beyond the experimental thing, we don't want to get a harness plugged on the side of our head when sitting in the car or get equipment as we were in an aircraft so the reply could be to locate whole the stuff in the middle mirror, behind the chrome surface.

if some of you want to play with the code here it is

http://www.r-kraft.com/motempl2.c

it won't work as is, you need to download the Intel's library from sourceforge.

Just thinking out loud... after seeing your code track hand movements...

Wouldn't it be fairly straightforward (LOL) to read sign language?

You can have a circle/zone/whatever follow each finger and two parts of the palm. Logic dictates that the 7 entities can be only arranged a certain way and then use a LUT (or similar) to match the hand shape/motion to sign language?

JAT

This would probably be nice but this is not what i'm actually doing and i don't even know how i could do that at the moment but if you want it's just up to you to download the opencv library and give a try to this idea

They are working at the CMU on this axis multimodal recognition and especially turned towards the signs language for people with disabilities.
This would add life comfort for those sadly hit by the life people.

found a good candidate, an old webcam i found between things to scrap.

I'm installing needed modules and i'll give it a try in the next hours to check how the IR things works, following pupil as i think or not