RE: [SeattleRobotics] Re: PIR Sensor Modification
- Seems like someone would want to rasterize the data coming from a PIR, much
like from a camera. Then CV could process it, and define little rectangles
to track the desired object (once trained). Saw a good demo at the
HBRobotics club meeting last night!
It was funny, sometimes the camera would ketch a glimpse of Wayne, and would
stop tracking the speaker, and follow Wayne! Apparently the face
recognition algorithm liked Wayne's face better ?!
> -----Original Message-----
> On Behalf Of Dennis Clark
> I can see this working. It doesn't matter whether the sensor moves and
> sees a change in heat or the person moves and the sensor sees a change
> in heat. The sensors are usually tuned to human body temperature so
> just a background heat difference wouldn't trigger it (unless it was at
> body temperature). If you pointed the sensor, then if something was
> there, you'd trigger it, the motion caused the differential. I wouldn't
> think that it'd be all that accurate, but it should still work. What an
> interesting project!
- --- In SeattleRobotics@yahoogroups.com, "David Buckley" <david@...> wrote:
>I favor your scenario over Peters. :)
> Surely a non scanning PIR on a robot creeping towards you is exactly the same as you creeping towards a non scanning stationary PIR operating your kitchen lights. As long as something moves and the animal image is big enough and crosses a boundary then the PIR wiil see it.
> If the robot is heading for somone then eventuall the image will grow to a size which will cross a boundary.
>I wish I knew. Supposedly, frogs can only see moving objects [or at least they only "respond" to such]. Supposedly they starve in the midst of piles of dead flys [those old time researchers were kind of whimsical].
> A bit like a frogs vision?
> Hmm a question Dan, when a frog is moving about does its vision system switch into a different mode so that it can see staionary objects?
However, my guess is that even frogs have a primitive form of "efference copy" such that self-generated movements inhibit the systems that normally respond to external movement. Otherwise, the frogs would be continually triggering their own predator-prey behaviors whenever they so much as twitch.
In olden times, there was a lot of neurophysiology work done on the frog/toad optic tectum [the major predator-prey detection area, eg David Ingle, J-P. Ewert, Jerome Lettvin et al], but a few studies also showed there was an inhibitory loop involving the tectum and diencephalon [primitive thalamus], but I don't remember the details.
So, the basic connections are there for efference copy/etc.
[also, no reason not to make up something that sounds reasonable and stick it on a robot :)].