Main Page | Report this Page
 
Loading..
   
Science Forum Index  »  Image Processing Forum  »  moving object same color of the background +...
Page 1 of 1    
Author Message
...
Posted: Thu Jun 26, 2008 11:45 am
Guest
Hi!
I have a few problems with the background subtraction. I hope somebody
can give me any hints.

1)I have a stable camera recording at 60fps an animal walking in the
sagittal plane in a indoor area (the background is always the same).
I'm lighting the area using a few tungsten lamps (working at 60Hz).
Despite that, analyzing the frames one by one I can still notice very
small changes of the bright (flickering). Is that normal?

2)I use an algorithm for the background subtraction found at
http://staff.science.uva.nl/~zivkovic/DOWNLOAD.html
It's working quite well (I almost get the perfect shape of the animal)
but I still have problems with the extraction of a leg of the animal
(that has, more or less, the same color of the background). Anyone can
suggest me what I can do to improve the algorithm? Especially for the
parts that are similar to the background..

3) for every video I have to set different parameters to make the
algorithm working well (due to light changes). But I would like to
avoid that in order to use it in semi-real time application.

I even tried to calculate the mean of the background (~50 frames) and
then subtract it to every frame but no good results.

Thanks in advance!
Back to top
ImageAnalyst...
Posted: Fri Jun 27, 2008 5:22 am
Guest
On Jun 26, 5:45 pm, mariobard... at (no spam) gmail.com wrote:
Quote:
Hi!
I have a few problems with the background subtraction. I hope somebody
can give me any hints.

1)I have a stable camera recording at 60fps an animal walking in the
sagittal plane in a indoor area (the background is always the same).
I'm lighting the area using a few tungsten lamps (working at 60Hz).
Despite that, analyzing the frames one by one I can still notice very
small changes of the bright (flickering). Is that normal?

2)I use an algorithm for the background subtraction found athttp://staff.science.uva.nl/~zivkovic/DOWNLOAD.html
It's working quite well (I almost get the perfect shape of the animal)
but I still have problems with the extraction of a leg of the animal
(that has, more or less, the same color of the background). Anyone can
suggest me what I can do to improve the algorithm? Especially for the
parts that are similar to the background..

3) for every video I have to set different parameters to make the
algorithm working well (due to light changes). But I would like to
avoid that in order to use it in semi-real time application.

I even tried to calculate the mean of the background (~50 frames)  and
then subtract it to every frame but no good results.

Thanks in advance!

--------------------------------------------------------------------------------------------
Mario:
If your main requirement is to segment out the animal from the
background, then I think you should use chromakeying, like what
weather foreacasters use. If you can, you should make a fairly
uniform background using a cyclorama ("cyc"), also known as "seamless"
or "sweep." For example, here's some links:
http://procyc.com/product.php?intID=9
http://en.wikipedia.org/wiki/Cyclorama_%28theater%29
http://raykophoto.com/?page_id=23
http://www.google.com/search?hl=en&q=chromakeying
But you'll still have a problem if the color is the same color as your
background, because things like chromakeying have you put in some
tolerance for how close the color can be to be considered as
background. But such a uniform background will surely be better than
some cluttered background. Then you could always overly your animal
onto whatever background you want later (or in real time like TV
weather forecasters).

Regarding lighting, yes you will have flicker using such lights
because incandescent and fluorescent lamps that run off the line
voltage will flicker 120 every second (once on the upper part of the
sine wave and once on the bottom part). Phosphors and thermal inertia
(filament tends to stay heated) reduce the problem so you don't go all
the way down to zero when the voltage crosses the zero line, but you
still have a variation on the light output. the worst case is where
your light frequency is a multiple of your video capture rate, like in
your case - lucky you. This can introduce flicker or hum bars in your
video which runs at 29.94 frames per second (NTSC) or 60 fps in your
case (this is a manifestation of "aliasing"). You can use DC
lighting, such as LED's, or high frequency fluorescent lamps (which
run at kilohertz rates) to eliminate the problem. Or you can reduce
your speed so that you are capturing more light cycles per frame of
video so that the light variations will average out to a more constant
value. Sounds like you don't want to do that because you are
capturing at high speed, presumably to show the motion of the animal's
legs in slow motion. So in your case, you need to use either DC
lighting or high frequency fluorescents. You can also ask in
sci.engr.lighting.
Regards,
ImageAnalyst
Back to top
 
Page 1 of 1       All times are GMT - 5 Hours
The time now is Mon Oct 13, 2008 2:25 am
Science Forum Index

Powered by phpBB & ML | Disclaimer



Contact Us