Detection of Submarine Periscopes in High Sea Clutter

DESCRIPTION:
Under a Navy STTR project (Contract No. N00014-05-M-0237, Topic No. N05-T013), Virtual EM Inc. developed Space-Time Adaptive processing (STAP) techniques using machine learning algorithms for eventual implementation in critical radar systems. Virtual EM has successfully demonstrated the feasibility of using Support Vector Machines (SVMs) for detecting targets in the presence of clutter and interfering sources. Of particular interest to Navy was detection of submarine periscopes in rough sea and identification of small boats with malicious intent in littoral waters. A set of measured clutter radar data was obtained from NAVAIR and a synthetic Gaussian target profile, along with six-to-eight uniformly distributed interference returns, was inserted into the data sets. The training was done with the HH and VV polarized radar data sets while the HV-polarized data set was used to test the performance of the trained Support Vector Classifier (SVC). The accuracy of the SVC was demonstrated to monotonically increase with the Signal-to-Clutter Ratio.

SBIR TOPIC #: N05-T013 (Phase I)