[Lab] Depth Perception
tom.i.burns at gmail.com
Mon Aug 22 15:02:23 EDT 2011
a bit of math.
average distance between eyes = 6cm = 0.06m (women = 5.5cm average, men =
football field distance ~= 100m. (give or take depending on which league,
cloud height = depends on a lot of factors but lets say 2000m which is what
wikipedia suggests is "mid range clouds" distance.
scale factor = 100 / 0.06 = 1666.67
so at a football field's pupillary distance a 2000m high cloud would appear
2000/1666.67 = 1.2m.
For where to cross the eyes I would guess a comfortable point would be
around 1m in front of your eyes, so around 1666 m in the sky.
Practically speaking aligning the cameras would be a nightmare.
On Mon, Aug 22, 2011 at 2:46 PM, <aaron at ottawarobotics.org> wrote:
> My thought was that the cameras won't be able to rotate a center point like
> our head is able to. Tying the camera rotation to the head movement instead
> of eye movement (which is what the cameras will actually be able to do in
> that scenario) might feel unnatural. Of course, tracking eye movement
> instead of head movement is a much harder thing to do, particular when your
> eyes will be covered with LCDs or something along those lines.
> On Mon, 22 Aug 2011 14:41:53 -0400, Darcy Whyte wrote:
> Um, if you moved your eyes (and the scene), it would be a strange feedback
> loop since your eyes are also looking at the scene.....
> I suspect that if the rates of movement were set right it might work out
> for matching your head movement to the sensation.
> You want to save the eyeball movement for moving around the screen.
> On Mon, Aug 22, 2011 at 2:38 PM, <aaron at ottawarobotics.org> wrote:
>> I think that tying the position of the cameras to your head movement
>> would cause vertigo or something similar, due to the camera rotation not
>> matching what your head motion is doing.... what would be better is some way
>> to track your eye movement and position the cameras that way. That way the
>> camera motion would resemble what your eyes are trying to do.
>> On Mon, 22 Aug 2011 13:35:28 -0400, Darcy Whyte wrote:
>> Why not use a matrix of Web cams? By the time you figure out all the stuff
>> like how to point the cameras with servos to keep them aligned you may as
>> well have a few of them. :)
>> You could put an accelerometer on your head so you an look around.
>> But unless the objects are vary far you'll want to work out how much to
>> tow them in (when you look far and close your eyes cross inwards when seeing
>> That's the only thing that isn't really obvious. How to decide where the
>> two cameras cross over.
>> I guess you could move your head forward and back and the accelerometer
>> could control that?
>> On Mon, Aug 22, 2011 at 1:28 PM, Michael Coyle <mirk44 at hotmail.com>wrote:
>>> Did anyone see today's xkcd?
>>> I need to try this!
>>> Lab mailing list
>>> Lab at artengine.ca
>> Lab mailing list
>> Lab at artengine.ca
> Lab mailing list
> Lab at artengine.ca
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Lab