Multi-Agent Methods: An Example of an Architecture and its Application for the Detection, Recognition and Identification of Targets - Information Fusion in Signal and Image Processing - page 217

which allows us to specify the names of the concepts we want, the types of methods
necessary to their analysis and to calculate in semi-automatic mode the parameters of
these methods. In order to do this, the areas of interest that represent each concept
(road, vehicle, etc.) are selected in an interactive fashion in the image.

At each new image, this knowledge base is updated based on the information gath-
ered by the agents. This update currently concerns the method parameters. A prospect
for the future is of course to extend this “learning” to the choice of methods.

10.5.2. The world model

The information gathered by the agents over time progressively adds detail to the
world model. A representation of this structure is shown in Figure 10.7.

concept info
identifier

image info
identifier
reference image
image(s) inter
image hough,
image seg, etc.
image size
concept list

sample list
support list

support info
identifier

estimation

area
type
Nb pixels
sample list

sample info
identifier
identifier parent
type
type parent
Nb pixel
estimation
statistic info

statistic info
mean
standard deviation

variance

link between 2 elements
of a list
composition relationship

maximum
minimum

Figure 10.7. Structure of the world model: for each image, we have access to the list of
concepts, of supports and of the associated statistical parameters. A concept
is comprised of several supports, which are themselves comprised
of samples characterized by a list of statistical information