Page 354 - Concise Encyclopedia of Robotics

Vision System
The moving image, received from the camera and processed by the
circuitry,contains an enormous amount of information.It is easy to present
a robot controller with a detailed and meaningful moving image.It is more
difficult to get a robot to “know” what is taking place in a particular
scenario, based on the visual data it receives Processing an image, and
getting all the meaning from it, is a challenge for vision-system engineers.
Vision and Artificial Intelligence
There are subtle things about an image that a machine will not notice unless
it has an extremely advanced level of artificial intelligence (AI). How, for
example, is a robot to “know” whether an object presents a threat? Is that
four-legged creature a big dog, or is it a tiger? How is a robot to know the
intentions of a moving object, if it has any? Is that biped creature a human
being or another robot? Why is it carrying a stick? Is the stick a weapon?
What does the biped want to do with the stick? It could be a woman with
a closed-up umbrella, or a boy with a baseball bat. It could be an old man
with a cane, or a hunter with a rifle. It is simple for a human being to tell
the difference and to gauge the appropriate behaviors for dealing with
any of these situations; programming a robot to have the same level of
judgment is exceedingly complex. You know right away if a person is
carrying a jack to help you fix a flat tire, or if the person is clutching a tire
iron with which to smash your windshield. How is a robot to know things
like this? It would be important for a police robot or a security robot to
know what constitutes a threat, and what does not.
The variables in an image are much like those in a human voice. A
vision system, to get the full meaning of an image, must be at least as
sophisticated as a high-level speech recognition system. Technology has
not yet reached the level of AI needed for human-like machine vision and
image processing.
Fortunately, in many robot applications, it is not necessary for the robot
to “comprehend”very much about what is happening. Industrial robots are
programmed to look for certain easily identifiable things. A bottle that is
too tall or too short, or a surface that is out of alignment, or a flaw in a
piece of fabric, are easy to detect.
Sensitivity and resolution
Two important specifications in any vision system are the sensitivity and
the resolution.
Sensitivity is the ability of a machine to see in dim light,or to detect weak
impulses at invisible wavelengths such as infrared (IR) or ultraviolet (UV).
In some environments, high sensitivity is necessary. In others, it is not
needed and might not be wanted. A robot that works in bright sunlight

