|
|
ATR: Automatic Target Recognition
We present a new approach to feature-based Combat target Identification (CID) for RF sensors, such as
Synthetic Aperture Radar (SAR) and Inverse Synthetic Aperture Radar (ISAR). Our approach utilizes a
comprehensive set of localized features and their respective relations as extracted from the RF images.
It uses real-time inexact matching algorithms that can significantly improve the CID performance under
challenging weather conditions and operational scenarios. The approach captures a comprehensive set of
features and relations onto a hierarchical Attributed Relational Graph (ARG), as a unified representation
scheme to handle the diversity of features and relations extracted from the RF-sensor data.
Our ARG representation is an image-based, “target centered” representation scheme. Moreover, this representation
scheme will also prove suitable for capturing feature sets that represent previously detected false
alarms, so they can be stored in a false alarm database. Our approach includes a suite of new and
innovative algorithms for the efficient processing (e.g., inexact matching) and manipulation of the ARG
representation of sensor data as well as target and false alarm models to perform indexing, inexact
matching, evidence accrual and search space management functions to further improve the capability of CID
systems in dealing with partial information, concealed targets, false targets, and high-cluttered backgrounds.
|
