RE: [Algorithms] Motion tracking from a live video source

[Willem de B. <wd...@mi...>]

Hi,

=20

What exactly are you trying to track? Is it faces, or human body
movement? The literature

out there is vast, and a best solution depends on the context (ie., is
the camera moving

relative to the object; what kind of object are you tracking).

=20

Cheers,

=20

---

Willem H. de Boer

Homepage: http://www.whdeboer.com <http://www.whdeboer.com/>=20

=20

=20

________________________________

From: gda...@li...
[mailto:gda...@li...] On Behalf Of
Alias
Sent: 08 October 2005 14:45
To: gda...@li...
Subject: [Algorithms] Motion tracking from a live video source

=20

Hello all,

At the moment, I'm trying to figure out a couple of things relating to
webcam motion tracking.

Firstly - as I see it, there are two basic approaches:

1) Look for a certain colour, say, orange, and track it by checking for
pixels that match it, within a certain threshold.

2) Look at the difference between frames using difference blending, then
compare the bounding rectangles of the current frame with the previous
frame.=20

I've had some success with the first approach, but the the results seem
to vary wildly depending on the lighting - it works really well in low
light, and not so well in brighter environments.=20

So now I'm considering the second approach. However, what I'm trying to
figure out is the best way to get a movement vector from the bounding
rectangles.

Should I compare the current frame to the previous frame and then try to
interpolate a motion vector from that? Should I use the centre point of
the rectangles, or the corners, or is there some other technique that I
should know about?=20

Any info would be much appreciated,

Thanks in advance,
Alias