DPRG
DPRG List  



[DPRG] webcam robot vision limitations

Subject: [DPRG] webcam robot vision limitations
From: Jeff Koenig koenig.jeff at gmail.com
Date: Sun Jul 1 12:16:07 CDT 2007

dpa wrote:

"...jBot only returns 4 samples (like 4 frames) per second, less than
ideal, but it turns out that is enough for fairly robust navigation.
I suspect the same is true for video
processing: 30 or even 15 fps is not necessary, and 4 fps might do."

What little I remember of the Nyquist-Shannon theorem regarding
sampling rate says that you only have to sample at a sufficient
frequency to prevent aliasing (the sampling rate needs to be at least
double the maximum signal frequency to reconstruct the signal).  So,
if your robot's perceived surroundings are moving at a fairly slow
speed, a slow sampling rate might be perfectly fine.

Of course, if your robot or its obstacles are moving very fast, you
probably need faster sensor sampling.  I think human eyes sample at
between 10 and 50 hertz, which is apparently sufficient for us.  And
when we don't get many visual clues (like from intermittent low-light
situations), what do we do?  We slow...way...down.

More information about the DPRG mailing list