Image Processing Reference
In-Depth Information
The exploration graph associated with this segmentation is shown in Figure 10.15.
road objective
vehicle objective
first agent created
in the image
n
o
n
o
n
o
the first letters indicate
the agent type:
C: concept agent
L: localization agent
Dt: detection agent
F: focusing agent (index by a type: o on; n next to)
Dc: decision agent
the latter specifies the objective:
V: vehicle
R: road
indicates the probability
for a support to be a
member of the vehicle
concept
Figure 10.15.
Agent graph related to vehicle detection when two focusing rules are specified.
The type of the initial objective is “road”. The system searches
for vehicles “on” roads or “next to” roads
The first agent created is the Concept agent. It initializes the system and activates
the spatial exploration by creating a localization agent, which searches for elementary
areas of the image satisfying all of the localization clauses. Based on three extracted
samples, three detection agents are given the task of extracting the road supports. We
then note that the focusing branches related to the road objective split in two. This
indicates that two focusing agents are activated when a road support is found. The
system actually focuses “on” the road support, which is completed by also focusing
“next to” road supports.
Five kinds of agents are required in this case: the Concept, Localization, Detection,
Focusing and Decision agents. They can be designed with two possible objectives:
road and vehicle. The segmentation result is shown in Figure 10.16.
Four vehicles are extracted. Three are located on one of the road supports and the
latter is located next to that same support. The system detects all of the vehicles in
Search WWH ::
Custom Search