So, if one buids a system according to the specifications given till now let's see how it is going to see.
The cameras are fitted with a motor, controlled through software.
The two motors are controlled in such a way that they will not be able cross their acute angle simultaneously.
The computer it is fitted to has some memory, but is void. No database yet, each and everything has to be done by the system
itself. When the cameras open their eyes what they see is a lot of colors all around, which they are unable
to interpret or understand. So, how does it kick start?
First of all what
we need to concentrate on is, how is it going to see a single object, when it has two eyes at an offset? Correlation
is a solution, but it is not so straight forward. You are going to run out of real-time if you start correlating the entire
image. Our cameras are equipped with a good focussing capability which is going to decrease the correlation time.
There are two things
that help to achieve this, the angle of convergence of the eyes and focus. By this time a question would have arised
in your mind asking you, why can't focus information alone be used, if you are depending on it for depth information? We are
definitely relying on the focus information for depth but not for continuity. This is because focus is not very sharp, and you cannot expect each and every point along your line of sight to have a different focus. There is
a small value delta(depending on the actual depth it may be a few mm, cm, m, etc) over which the focus value remains the same and
within which the various points are not distinguished. The angle of convergence of both the eyes combined is unique
for each and every point in the 3D space. Then why not use only this information to perceive depth? Definitely possible! but
if you have understood what I was trying to explain about focus till now, the answer should be with you by now. The value
delta I was talking about has something to do with correlation and fovea vision. This is where each of the concepts discussed earlier converge. When you look at an object, depending on its depth your eyes
is going to focus. This focus information is going to tell you that depth(+/-)delta/2, is where you are seeing. Now just correlate
your foveal area and you will be able to find a perfect match if one is present. If it is binoculor friendship, then anyway one of the images is suppressed and only the other is seen.
|
|
