FreeTrack Forum

Welcome, you're not connected. ( Log in - Register )

FreeTrack Forum > FreeTrack : English Forum > General Discussion > Other way too manage input information ...

RSS >  Other way too manage input information ...
themacboy #1 08/05/2008 - 17h21

Class : Apprenti
Posts : 4
Registered on : 08/05/2008

Off line

I see this system always works with high contrast image that you get when iluminate a webcam with a led. But it is not possible to use color and not light? for example:

I was in my room, there is a normal iside light. In my room there's nothing with green color, then if i use a green color circle paper it is not possible to modify the algorithm of freetrack to check for green color objects?


oh sorry for my bad english !
didja #2 08/05/2008 - 20h05

Webmaster (admin)
Class : Webmaster (admin)
Posts : 1043
Registered on : 09/07/2007

Off line Www

That is a really good idea, but I think the tracking wouldn't be as accurate as it's with the actual system... But if it would be possible, it would allow tracking without any LEDs, like a few piece of paper...

Rémy
Posted Image
Deimos #3 10/05/2008 - 22h30

Class : Beta Tester
Posts : 120
Registered on : 07/11/2007

Off line

It is possible, can be quite accurate and isn't too hard to implement. I made a little program doing color recognition and driving a robot to follow an object of defined color as a university project a few years ago. The color recognition part works like a charm, even with non-uniformly lit objects, providing stable and reliable output ("blobs" of recognized color - pretty much what you see in FT preview window). I did the recognition by converting image colors from RGB colorspace to HSL colorspace. From there it's rather easy - you just need to specify minimal and maximal values of hue, saturation and lightness parameters and check if color of a pixel is between these values.
But the problem is that the camera has to actually SEE the color, so there has to be plenty of light, and rather long exposure times are needed (so lower frameretes can be achieved) - and that will significantly decrease tracking responsiveness.
Kestrel #4 11/05/2008 - 07h57

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

*You're limited to the colour green because the Bayer filter in digital cameras has half the resolution for red and blue.

*The points would need to be bright green balls and be fairly large to account for the small surface area of the ball that passes filtering.

*The location and colour of lighting (incandescent is yellow, fluorescent is green etc)  and even the colour of the room can influence the way the tracking points are lit.

*Camera brightness, contrast, saturation and white balance need to be carefully managed to get the right colour range, something you don't need to bother with for tracking bright objects.

*Cameras have a small aperture and can't see well in normal room lighting when operated at their max frame rate.

*If you need to use dedicated illumination for the points you're better off using an infrared lamp and retroreflective material.

*Professional mocap solutions use illuminated markers.
Deimos #5 11/05/2008 - 19h39

Class : Beta Tester
Posts : 120
Registered on : 07/11/2007

Off line

Well, i can't agree with you on the green color, Kestrel. It is possible to track all colors. However as you said - it would require quite a lot of setting up: white balance, exposure, gain, color thresholds for color recognition, etc.
I recorded my program i mentioned earlier in action:
http://www.youtube.com/watch?v=SdqmYjb9jc4 (make sure to click "original movie size at youtube, otherwise it will be badly stretched).
The room is lit with combination of daylight (a cloudy day) and incandescent light. The camera is working with exposure allowing 30FPS (i got  less in my program because of rather crappy point recognition algorithm and because i have some sleep() commands in the code to make room for communication routines).
The circular dial sets up minimal and maximal Hue of the target color, the top bar sets saturation threshold, and the lower bar sets lightness thresholds.
As you can see, it can track green objects quite well, but there are no problems with blue or orange objects as well.

But as you said - it's not very practical and using LEDs is much better. But mow that i think about it - how about some basic color filtering to make FT recognize a LED only in a given color - for example a green or blue LED should be easily picked out from background even with some parasite background light. If you're doing filtering in RGB colorspace, then it shouldn't be that hard to process only one of the channels instead of all 3?
Personally i use IR modded webcam, but this might be a good tradeoff for people who don't want to modify their cameras...
Edited by Deimos on 21/07/2008 at 18h25.
themacboy #6 12/05/2008 - 22h48

Class : Apprenti
Posts : 4
Registered on : 08/05/2008

Off line

Oh nice explanation.

The reason because i suggest this technic is to avoid the use of any elament that needs power (AA, AAA, sr44, etc), and avoid to full the room with IR light.

This is because I think this system is better.

Thank's a lot, again and again, for all clear explanations I see this complexity of this system.
didja #7 12/05/2008 - 22h59

Webmaster (admin)
Class : Webmaster (admin)
Posts : 1043
Registered on : 09/07/2007

Off line Www

Hi :)

themacboy @ 12/05/2008 - 22h48 a dit:

and avoid to full the room with IR light.



You know, that's not really a problem since IR light is... invisible !  :D

Rémy  ^^  :)
Posted Image
themacboy #8 13/05/2008 - 01h12

Class : Apprenti
Posts : 4
Registered on : 08/05/2008

Off line

hahaha sure but I use IR for other jobs in my room ... think about it hahaha

didja @ 12/05/2008 - 22h59 a dit:

Hi :)

You know, that's not really a problem since IR light is... invisible !  :D

Rémy  ^^  :)

Deimos #9 13/05/2008 - 02h02

Class : Beta Tester
Posts : 120
Registered on : 07/11/2007

Off line

Okay, you sure got a point with avoiding batteries - i use a passive system (illuminator + reflectors) myself.
But if you're concerned about illumination for a passive FT setup jamming your TV remote, IRDA, IR-controlled toy helicopters, or other communication methods using IR, you don't have to worry - they all use IR light modulated with a certain frequency, and receivers are adjusted to pick up only this single frequency signals. FT illumination is constant light, so it's filtered out by receivers' filtering circuits.
Actually, light from normal AC lightbulbs used for room illumination causes way more IR interference than any FT setup, because:
-besides visible light, they also emit lots of infrared
-their light is modulated with frequency of current (50Hz or 60Hz, depending on country), while FT illumination is constant light.
-unlike FT, its light floods the entire room, while IR illuminators usually work in a rather narrow cone (~50degrees).
Besides, if for some reason you don't want high-intensity IR light in your room, you can use an active system and power the LEDs from USB port, or another external power supply - additional cable isn't much of a problem when using headphone clip, since it can be attached to headphone cable.

 >  Fast reply

Message

FreeTrack Forum > FreeTrack : English Forum > General Discussion > Other way too manage input information ...

 >  Stats

1 user(s) connected during the last 10 minutes (0 member(s) and 1 guest(s)).