The Naked Scientists

The Naked Scientists Forum

Author Topic: infrared light for human shape detection - need some general help  (Read 5086 times)

Offline FabianB

  • First timers
  • *
  • Posts: 4
    • View Profile
Hello,
my name is Fabian and i'm writing my master thesis of computer science at the University of South Australia.
Im working at a project concerning the llumination of real 3D models with multiple projectors.

I need some general help and information about IR and i hope, you will help me..

My goal is to detect the shadows on the model cast by users stepping in front of the projector. Three different approaches dealing with IR might be possible..

1) Illuminating the whole scene with IR-light and detecting the IR shadow of the user on the model.

2) Is it possible to detect the user via thermography in a multi-projector illuminated environment?

3) or do you have knowledge of any white paint color with special reflective attributes?
e.g. reflecting visible light in a normal way, but absorbing all IR waves? (that would be the easiest solution for me)
or a material that reflects visible and IR waves in a different way?

I ll appreciate any useful comments on this topic or further links to other forums/information..
thanks & cheers
Fabian


 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Heh - I'm sure your posting makes sense to you but I'm a bit confused by it.

Point 1) doesn't seem to require any comments.

Point 2) - I can't think of any intrinsic reasons why it shouldn't be possible to detect 'users'(?) via thermography in a multi-projector illuminated environment. But I don't think that's going to help you.

Point 3) - I probably wouldn't know of any special paint but are you sure you want to absorb all IR?  If the IR is absorbed at reflection, how are you going to detect it post reflection?  I'm not sure that checking for an abscence of something is a good idea here - you could get abscence of IR because of shadow but you could also get it if the projectors fail.
 

lyner

  • Guest
This is a novel position sensor  for pointing at the screen? You want the projected image to be unaffected but you want the screen to absorb enough energy to have a measurable increase in temperature?
Brilliant idea as it would mean no more expensive and vulnerable 'Smart Board' surfaces - just a passive screen.
You presumably want a cheap solution so a paint would be good. However, it is easy to produce dichroic   (http://en.wikipedia.org/wiki/Dichroic_filter)  filters which would pass IR onto a black surface beneath yet reflect visible light. Easy but not necessarily a cheap technology.
I have a problem with the idea of using multi projectors, though. Wouldn't this produce multiple shadows and introduce confusion? Or is this an attempt to eliminate the shadow of the user's body by having the sources 'near the wall'. The only spot on the screen where you would have a good shadow would be just under your finger, where all the shadows coincide. Smart idea.  But what sort of resolution would you get? Also, how many IR cameras will you need if you want to avoid masking by the operator's body?
It may be an idea to rush off and see what potential for patents there is. Or it may all have been thought of before. Perhaps you are way ahead of me, there.
I have a feeling that the operator could get a bit warm, using your invention?
Keep us up to date with the proj, please!
« Last Edit: 19/05/2008 12:58:17 by sophiecentaur »
 

Offline techmind

  • Hero Member
  • *****
  • Posts: 934
  • Un-obfuscated
    • View Profile
    • techmind.org
My goal is to detect the shadows on the model cast by users...

I think you need to clarify exactly what you mean by detect the shadows.

What are you trying to do with the shadow-data? Do you simply need to know that a shadow has been cast ("yes/no") are are you looking to infer position/shape? What do you want to happen when the body of the person casting the shadow obscures the camera view of the model?

This could be quite a difficult problem, but it's difficult to be much help without more specific questions.

Standard black-and-white "security" CCTV television cameras are very sensitive to infra-red. Also webcams etc if you remove the IR filter (at your own risk!) If you can obtain a visible-blocking filter (looks very deep red, like the plastic at the front of IR remote controls) and put it in front of such cameras you will get an IR-only image.
Are you relying on the IR from the projectors, or will you have a separate IR source? Modern "business projectors" use a discharge bulb which won't emit that much IR, and there may be inbuilt filters to block it (because the heat damages the LCD). There may still be enough IR for this application...?
 

Offline FabianB

  • First timers
  • *
  • Posts: 4
    • View Profile
Hi there!
thanks for your interest, i should have explained the problem more detailed than i did..
to make things clear i will post a part of my thesis plan, that will hopefully explain the basic problem:

Quote
Augmented reality comprises a wide range of different applications, techniques and input/output
devices. In general it supplements the real environment with additional, virtual information.

The spatial augmented reality (SAR), visualizes the virtual content directly on the surface of physical
objects. The information can appear on arbitrary surfaces or aligned to real 3D objects. Users of this
system are unconstrained from hardware devices such as head-mounted-displays. Projection-based
spatial displays use front-projection to support multiple users independently. The virtual information is
presented on a wide field-of-view and in high resolution. The shader lamps technique illuminates
diffuse white objects with multiple projectors, simulating different surface appearances.
While interacting with a shader lamps system, users will be casting shadows on the object, destroying
the intended impression.

Compensating these shadows with additional projectors will improve the usability of projection-based spatial displays.
The main goal of this thesis is to develop a prototypical shadow elimination algorithm for shader lamps.

This is an example for a shader lamps project:


 ------------------------------------------------

To reach this goal it is compulsory to detect the user shadow or the user shape !


APPROACH 1)
Former 2d approaches have used the infrared technique to distinguish the usershape from the surface of the model.
The algorithm works as follows:

Unlike the user, the surface is illuminated with (i suppose) near infrared waves!
NIR cameras, record the high IR contrast between the IR unilluminated user and the IR illuminated surface.
(This results in a detected user shape and methods can be started to compensate the shadow on the surface...)

The following graphic explains the setup for 2d shadow detection:


However, in a 3d environment with a movable user it will not be possible to illuminate just the surface and NOT the user.
Instead you have to go a vice versa way and illuminate the wohle scene with IR light. So the user will cast an IR shadow on the surface, that could be detected with an IR camera...
As techmind said, it would be better if the small IR radiation, emitted from projectors could be detected...
(But im not sure about this)



---------------------------------------

APPROACH 2)
This approach could be enhanced if it is possible to detect the "normal" IR radiance emited by humans AND if this irradiance is recognisable by cameras in a projector illuminated environment. Maybe it is considerable, that standard projectors filter out IR radiance...


---------------------------------------

APPROACH 3)
 
However, the best solution would be an IR absorbing white material !!!
The whole scene would be illuminated by IR light, but only the user would reflect it!



---------------------------------------
If you are generally interested in this topic, you can visit my homepage: newbielink:http://www.uni-koblenz.de/~buetz [nonactive]
(its more a private note page, than a representative page)

Anyway, thanks for your comments so far, maybe you have further contributions?


---------------------------------------

@sophiecentaur:

Im just as confused of your post than you were of mine ;)
hmmm.. i havent done any kind of invention... but it seems to me you did, with your interpretation of my post ;)
Pls..could you explain your thoughts more detailed.. it sounds interesting!

 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Ah - now I think I understand what you're aiming for.  If I'm not wrong, there are some similarities here with material mapping in 3D computer graphics, in that you're projecting surface details on to a 3D object, but the problem occurs with observers getting in the way of the projection.

It seems to me that you won't be able to use a simple projector to display static data on to the objects because there'll be no way to vary the output across the display to compensate for the shadow areas i.e. you'll need to project the surface details as a raster so that you can increase the brightness across the shadow areas to compensate for them.

If I were trying to design such a display system I think I'd do away with the IR bit entirely and instead, use a sort of crossed feedback loop.  To do this I would use cameras co-positioned with the projectors to view the results of the total projection.  The data from these cameras, which would include any shadows, could then be compared with what it should see if there were no shadows and use this to apply compensations.  In effect, each camera would actually be looking for shadows cast in the displays from the _other_ projectors and will be telling it's projector how to compensate for those shadows.

In your diagram, a camera co-positioned with projector P1 would be able to see Shadow2 cast in the output of P2 and adjust the output of P1 to compensate.  Similarly, a camera at P2 would see Shadow1 cast in the output of P1 and compensate for it.

I'm afraid that if the IR bit is mandatory to your project this probably won't help much.
 

Offline FabianB

  • First timers
  • *
  • Posts: 4
    • View Profile
@LeeE: You're completly right! (camera attached to projector an easier way)

BUT: ;)

Quote
"could then be compared with what it should see if there were no shadows and use this to apply compensations"

..you dont know what the camera SHOULD see... thats the point!
(you know what images are leaving the projector, but you dont know how they are observed by the camera...)
To be independent of the DYNAMIC projected images and what the camera sees, i would like to use IR waves!

I know the model is illuminated by IR waves, so i know how the model SHOULD be (all IR illuminated),
if the user steps in front of a camera, he is not IR illuminated!
And the image that should be is superimposed by the user! -> i know where the shadow will be...

-------------

 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Why don't you know what the camera should see?  You know what data you're projecting on to the object and you know, or should be able to find out, the characteristics of the object that you're projecting onto, but even if you can't I would have thought it would be relatively quick and easy to calibrate such a system.

For example, and assuming you're using an RGB projector, I'd start by illuminating the object with sweeps of colours and intensities to get, in effect, an intensity mask for the scene, for those colours and intensities from each of the cameras.  Once you've got that you should be able to predict what the cameras should see, regardless of dynamic content, because you should know what you're projecting.
 

lyner

  • Guest
It strikes me that medical imaging is where you ought to be looking for ideas. Whatever wavelength you are using, you  get a  set of shadows according to the direction of illumination and, if you are lucky, the processing can identify the image.

Re my first post ; totally at cross purposes but I read yours as a basis for a design for locating your finger (pointer) on a Smartboard - type screen without having any electronics in the screen itself. With multiple infra red sources, there would be many penumbras / light shadows cast on the screen but only one umbra, corresponding the dark shadow under your finger right near the screen. This could be detected with an array of cameras placed obliquely around the screen.  Cheap and robust; childproof, even.
There - you invented something without knowing it!
 

Offline FabianB

  • First timers
  • *
  • Posts: 4
    • View Profile
@LeeE:
i see, you are really into this subject! Youre ,again, completely right with your idea!
There is an existing 2d system that computes such an intensity-transfer-function between ougoing projector intesities and incoming observable camera intensities..
But this system is very slow and not capable of realtime rendering...

any other ideas or knowledge of special reflective materials?
or do you know where i can get information of this?

thanks,
Fabian
 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Sorry - I've no knowledge of special reflective materials or where you could get info on them.

I worked in IT for 30 years and at the same time pusued photography as a hobby until I eventually moved on to using 3D graphics s/w instead of cameras, so it's probably from that combination that any understanding I have comes.

I'm a bit surprised that such a system would be too slow, at least with today's tech.  Although the quantity of data in any visual system is going to be very high, requiring a lot of processing resources, it's a natural candidate for MPP (massively parallel processing) - each calculation shouldn't be very intensive, after all, you're just comparing the camera data with a calibrated mask to calculate a correction, but how many of them you have to do will depend upon the frame rate and resolution.
 

The Naked Scientists Forum


 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums
 
Login
Login with username, password and session length