A cornerstone of our ATR product is the capability
to perform in-house development, testing, and real time implementation
of powerful algorithms for target recognition. In the algorithm development
phase, the expected appearance of target vehicles in SAR imagery is modeled
from available data. The degree to which target images may be expected
to vary about this model, under operational conditions, is also quantified.
 
|
|
A statistical target model for one of the ATR's identification
algorithms is depicted along with a one foot SAR image of the target vehicle,
a T72 tank
|
Statistically robust match metrics gauge the level of agreement between
target models and unknown objects in new SAR imagery. These metrics are
derived from rigorous mathematical principles and are specifically designed
to perform well in the presence of target signature variabilities arising
from diverse sources such as articulating or rotating target parts; changing
background surfaces and vegetation; partial target obscuration; and attempts
at camouflage, concealment, and deception. |
