MSTAR Match

Customer: DARPA ISO

Agent: Air Force Research Lab, Mr. Joe Diemunsch

Program Goals:

The goals of our Match module effort are:

Develop effective Bayesian scoring measures that discriminate between similar looking targets in the presence of modeling and imaging uncertainty, while robustly handling confounding scene factors such as obscuration, layover, and background clutter.

Efficiently search over target state parameters, such as target pose.

Characterize residual mis-match.

Key Technology:

We are integrating a suite of feature matching algorithms that capture a wide range of target appearance properties:

Target and shadow shape matching for characterizing the geometry of the target.

Target signature texture pattern matching for characterizing radar scattering properties of the target.

Symbolic target part descriptions, such as gun barrels and turrets, for component analysis.

We have developed Bayesian metrics for evaluating hypotheses' matches over these features along with effective optimization strategies to maximize the match metrics over target state parameters.

Key Product:

We are producing a software package that evaluates the degree of match of two images, factoring out pose parameters. The product is useful not only for recognition applications, but for registration and change detection problems as well.

Points of Contact:
Dr. Gil Ettinger, Match Principal Investigator: (781)-273-3388, x334
Mr. Greg Klanderman, Match Software Lead: (781)-273-3388, x337
Dr. William Irving, Technical Support: (781)-273-3388, x285
Dr. John Wissinger, MSTAR Program Manager: (781)-273-3388, x239

Back to MSTAR Top Level Page

Last Modified 12/07/98