Robotics Reference
In-Depth Information
vision programs: the detection of edges, the creation of line drawings us-
ing those edges, employing models of various shapes and mentally rotat-
ing them, and comparing the different views of models with the created
line drawings.
Roberts' approach was taken further by Adolfo Guzman in 1968.
Guzman's program, which he called SEE, detected junctions of lines,
points where two or more lines meet, rather than the lines themselves,
and could analyze scenes without the use of pre-stored models of objects.
The program could “look at” a line drawing of a scene and identify all the
objects in the drawing, even if they were not all completely visible. The
goal of the SEE program was to divide a scene into images of distinct
three-dimensional bodies so that a program such as Roberts' could be
used to identify each of the objects.
Guzman's program employed heuristic rules about the junctions of
lines on the surface of an object and what type of regions on the surface
converged at each type of junction. This approach allowed the program
to determine when two regions in a line drawing both formed part of
the same object. After analysing the regions adjacent to a junction, the
program would be able to suggest heuristics that were appropriate for
deciding to which object a particular region belonged.
Figure 31 is one of eight types of junction employed in SEE, in which
two of the regions are linked. Guzman called this type of junction an
“arrow”.
Guzman represented each type of object by a collection of links, in-
dicating which regions are adjacent wherever a junction occurs in a line
drawing. For example, in the case of the arch shown in Figure 32, the
regions R1 and R2 are linked at two junctions, as are R1 and R3, whereas
regions R2 and R3 are linked at only one junction. The program's rep-
resentation of the linking relationships is shown in the lower part of the
Figure 31. An arrow showing the two regions of an object that are linked
Search WWH ::




Custom Search