Re: Inexpensive infrared edge/cutoff filters

[Stephen Intille <intille@xxxxxxxxxxxxx>]

> Could you explain more clearly? Why not just flood the room with
> broad band IR and then selectively pick out the two areas of
> interest to you using "notch" filters? Or, alternately, why not
> flood the room with IR emitted by two different kinds of IR "LEDs"
> (or IRLED's!). These have fairly narrow emission bands and are
> available for several IR wavelengths.

Here's a followup to my first post with more detail. 

Here's what I'm trying to achieve. This idea was proposed by Jim Davis
and will clearly work but I'm trying to implement it for low cost.

You can see a better drawing/explanation and one application built by
Jim using one screen at
http://vismod.www.media.mit.edu/~jdavis/Pat/pat.html

In my scenario, a person stands in a room with two screens. The idea
is to project IR light onto a screen behind the person and observe the
person from the front with an IR camera. Since the person blocks the IR
light, you get a real sharp silhouette that (using computer vision) can
be used to extract the image of the person (analogous to
blue-screening).  From one direction, you have a cheap b&w camera, c1,
with a visible light cutoff filter, filter-c1 (some cameras don't have
IR cutoff filters and so will detect the IR). Behind the screen
opposite c1, you have a set of about 6 little $150 commercial IR
emitters.

The problem is that this only works in one direction. In two directions,
unless the emitters-1 and emitters-2 are sending out distinct 
frequencies, the scheme fails (because the front of the person
will be flooded with light from the second emitter!)

                       Emitters-2
                    /               \
                   /                 \
                  /                   \
                  --------------------- screen
                                      |\ 
       filter-c1                      | \
        ----                0         |  \
        |   >              \|/        |   Emitters-1
        ----                /\        |  /
         c1                           | /
                                      |/

                             ^   filter-c2
                            | |
                            | |  c2
                             -

Consequently, what I want is a *cheap* solution to do the following: 

Emitters-1 send out a wide band of near-visible IR to the screen 1.
Emitters-2 send out a mostly non overlapping band of near-visible
IR to screen 2. Then filter-c1 filters for the band of IR from 
emitters-1 and filter-c2 filters for the band of IR from 
emitters-2. Each camera can extract the silhouette of the person.

You can use edge filters to do this, with a cutoff around 900nm,
but they are expensive and for each screen I have 6 emitters 
that have a 5x5in surface to filter. I can't even find filters
that are this large, let alone affordable. 

One great solution would be to make the emitters using IR 
LEDs with particular frequencies, but I can't find a source
for these. I'd need LEDs centered around 840nm and 950nm with
bandwidths as narrow as possible. 

Any ideas appreciated. 

Stephen Intille
intille@xxxxxxxxxxxxx
*
****
*******
******************************************************
*  To remove yourself from this list, send:          *
*         UNSUBSCRIBE INFRARED                       *
*       to                                           *
*         MAJORDOMO@xxxxx                            *
*----------------------------------------------------*
*   For the IR-FAQ, IR-Gallery and heaps of links:   *
*  http://www.a1.nl/phomepag/markerink/mainpage.htm  *
******************************************************