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 222

All of these agents are not present in the image at the same time. As soon as an
agent has completed its task, it asks the administrator to delete its information file
before terminating itself. The result of the segmentation is shown in Figure 10.12.

probability included between 67 and 74%
probability included between 63 and 66%
probability smaller than 63%

4 vehicles present

false alarms

detected vehicles

Figure 10.12. Segmentation of the vehicles supports. On the left, the initial image shows the
presence of four vehicles. On the right, the resulting segmentation is shown. The shade
associated with each support represents the probability of membership to the concept

This segmentation image displays the nine “vehicle” supports found in the image.
A probability of membership to the vehicle concept is associated with each support.
This probability is represented by a color. It is interesting to emphasize two important
results:
- the vehicles are detected with a probability of membership of over 67%. On
the other hand, these elements of information are drowned in a large amount of false
positives with probabilities between 63 and 66%;
- this population of agents shows us a decentralization that appears very early on in
the system. The control is conducted locally by each agent. Furthermore, the steps of
the process are distributed so as to divide the computational load between the different
processes, as shown in Figure 10.13.

decision
detection
localization
concept

time

Figure 10.13. Distribution of the computation times among the different agents: black
indicates the activation of an agent and gray indicates that it is running