Image Processing Reference
The temporal exploration makes it possible to associate additional information in
the context of an information fusion cycle. For example, in an infrared image, a vehicle
can be characterized by a hot area in motion.
Just as with the spatial exploration, each phase of the information fusion is con-
ducted locally by a type of agent. Therefore, two new agents have been specified:
the Decision and Movement agents. The Movement agent searches for the support of
moving objects while the Decision agent's goal is to make the most of the two supports
extracted by the Detection and Movement agents, respectively.
The Decision agent uses the information obtained to assess the need to continue
with the analysis or to stop. This makes it possible to either activate the recognition
process if a target has been detected, or otherwise to continue the detection phase
by activating the loop related to the spatial exploration. As a result, the hypotheses
expressed regarding the support's identity and kinematics, respectively, lead to differ-
ent actions on the part of the Decision agent (see Figure 10.5).
results of the
results of the detection agent
Figure 10.5. The different behaviors of the Decision agent: the hypotheses stated according to
the support's identity and kinematics provided by the Detection and Movement agents,
respectively, lead to different actions
At each new image, the system's initialization and the knowledge base's updating
protocol are under the control of the Concept agent. This update allows the system to
adapt itself according to the gathered information and to progress toward the knowl-
edge specified beforehand. Thus, the parameters of the methods, among other things,
can be modified, such as, for example, the validation range related to the image pro-
10.4.2. Inter-image control cycle
There are three problems to consider in the inter-image analysis cycle:
- using the image sequence;
- using the information gathered from the images;
- updating the knowledge base.