[Avery.Andrews 940717.1135]
(Bill Powers (940714.1120 MDT))
>Hey, what's going on here? Where are all these features coming from?
Like I said, it was pretty sketchy. I'll try to amplify this bit. I assume
that if people have become familiar enough with an object to use it
for navigation, they acquire and store in memory some kind of 3d
representation of it. What kind of, I'm not sure.
One possibility is a Marr-style 3d model (like a collection of parameters
for a 3d-graphics program like POV or Polyray (a shareware program for
386/486 machines that I got recently). Another would be a collection
of `2 1/2dD sketches' of the object from various viewpoints (a 2 1/2D
sketch is a picture supplemented with depth info). Or maybe both at
once, for example a crude 3D model, plus some more detailed 3D sketches
of the most familiar aspects of the object (what the hill looks like from
you back garden, etc.)
The internal representation of the object, whereby the person or organism
identifies it, specifies the distginguished vector that is used for
orientation. Note that it doesn't matter what vector is used, and the
issue of how people choose one is distinct from how it is used in
navigation. Recognizing the object requires mapping the model and/or
sketches onto the view that you actually see (via mental projection,
scaling and rotation operations), and these operations will among other
things give you a value from the angle between the distinguished vector
(or one of them, if the mental representation of the object specifies
several) and the vector from the object to you. For 3D sketches it would
particularly simple: its the angle whereby you have to rotate the sketch
to make it fit the view you see.
Does this help? I'm pretty sure that I have not in fact been subverted
by Chapman and Agre on this.
Avery.Andrews@anu.edu.au