Robotics Reference
In-Depth Information
of these factors and more make it difficult for a computer to determine
which lines belong to a single object.
Some Applications of Computer Vision
One popular technique in computer vision is to compare new images of a
given scene with old images, ignoring anything that has not changed and
announcing when something was different between the earlier and the
later images. This task is of great importance, particularly in the realm
of security where noticing when a car starts to move or when a person
suddenly changes direction can provide clues about possibly illicit inten-
tions. Some advanced vision systems rely on having more than one image
to view at the same time, in the way that humans benefit from having two
eyes. Stereo vision systems employ two cameras that are positioned some
distance apart, as with human eyes. If a common feature in an object can
be identified in the images from both cameras, then the position of that
feature can easily be determined by means of triangulation—calculations
based on the distance between the cameras and on the angles of sight
from each camera to the feature.
Being able to see something is of little benefit if your brain cannot
make use of the visual information. You might be driving along a road,
seeing a car in front of you, but if your brain does not notice that the
image of that car is getting larger and larger at a very fast rate, then you
are likely to have a disaster on your hands. Modern computer vision
systems are often used in conjunction with software that employs other
aspects of AI technology, allowing, for example, for the development of
cars that are able to steer themselves safely along a road, or of systems
that recognize and interpret facial expressions. Even a “simple” task such
as determining that the black squares on a chess board are just that, part
of the surface and not holes in the board, requires a sophisticated vision
system.
Experience and general knowledge also play their part in enabling us
(and computers) to perform recognition tasks. For example, if a small
part of a scene is being hidden from view, such as when part of the
road on which you are driving is obscured by snow, your experience and
general knowledge tell you that, most likely, there is roadway beneath
the snow. By using such experience and knowledge Artificial Intelli-
gence programs can fill in the gaps, the missing parts of an image, to
