Image Processing Reference
In-Depth Information
10.1.1.
The application context
The objective of this work is to find the information regarding the target in an
image environment characterized by several levels of complexity:
- unpredictable context: the elements of the observed scene and the targets present
are not known beforehand;
- variety of possible scenarios: a scenario corresponds to an observation phase of
several targets in the considered environment. The scenarios are constantly evolving
in terms of luminosity, possible masking, shadows;
- multi-target application context: when processing a sequence of images, a target
can enter the sensor's observation field at any time;
- heterogenous observations: the considered targets are ground vehicles, aircraft
or boats. Depending on their positions with respect to the sensor, their aspects and
their time signatures are highly variable;
- incompleteness and imprecision: the information provided by the sensor is
tainted with imperfections and gives a partial interpretation of the scene.
We have chosen to limit the context of this study to the tracking of military targets
in a non-structured (non-urban), outside environment. Depending on the position of
the camera with respect to the target, we considered three different missions:
- “ground-to-ground” missions: the camera carrier is placed on the ground and
observes a scene on the ground; the possible target maneuvers in the observed space;
the sensor can be fixed or mobile; masking phenomena, in this configuration, are very
common and often caused by the presence of vegetation;
- “air-to-ground” missions: the sensor is embedded in an airborne carrier (plane,
helicopter). It is mobile and observes the scene on the ground; blocking is caused
mainly by vegetation-type masking structures;
- “air-to-air mission”: as before, the sensor is embedded in an airborne carrier, it is
mobile and observes part of the sky; target masking is caused mainly by the presence
of clouds, or by the possible presence of other targets.
Depending on the type of mission, the objectives, the target classes and their behav-
ioral characteristics vary. In a surface-to-air mission, the targets are planes. In an air-to-
ground or ground-to-ground mission, the focus is more on boats and ground vehicles.
This specification provides an element of information regarding the target's possible
class.
In the context of this study, we worked with images provided by infrared and vis-
ible sensors, acquired during surface-to-air or ground-to-ground missions (fixed sen-
sor). However, the system's design allows the integration of new sources of knowledge
and new algorithms adapted to other types of sensors such as the radar or other types
of mission.
Search WWH ::
Custom Search