All-Aspect Maritime Automatic Target Recognition

ISAR imagery has a long history of being used for maritime classification via airborne platforms.  Classification has been most effective when using ISAR images in which signatures from physical ship structures (deckhouses, masts, antennas, etc.) are separated in range and extended from ship hull signatures in the Doppler dimension (profile views) or when port and starboard features are separated in Doppler (plan views).  These views have required collection from  quartering angles (aspects significantly away from broadside and end views).Ship classification from ISAR imagery at broadside viewing angles is more challenging as mast signatures are overlapped or shadowed and the position of these signatures along the ship length is not measurable.  Ship identification is also difficult when using frames with minimal Doppler extent in which the masts may appear as peaks in radar cross section but not as extended Doppler signatures in the imagery.RDRTec has made significant advancements in the automated classification of ships using real time feature extraction from ISAR imagery. RDRTec’s Automated Combatant Vessel Classifier (ACVC) can classify ships to fine naval class level using ISAR data collected at long ranges. ACVC is actively being integrated into tactical programs for AAS, Triton, P-8A, and in discussion for transition to MH-60R, Fire Scout MQ-8B/8C, F-18, and MQ-25. ACVC is capable of classifying on a wide range of frame types beyond just pure profile and pure plan views.RDRTec proposes to address the challenge of aspect independent classification through the expansion of ACVC incorporating rotator detection and characterization, additional RCS feature detection, and aspect-dependent feature weighting.  This program will evaluate the effectiveness of ACVC at a range of target aspect angles and incorporate features and techniques to enable satisfactory classification at near broadside viewing angles. RDRTec will evaluate the improvement in ship classification for broadside ship geometries with the incorporation of signatures arising from rotating antennas. RDRTec has developed tools to detect and characterize micro-Doppler based signatures associated with rotating antennas, and indicated how such information could aid in classification, but the focus of these studies was on quartering aspect angles (not broadside) and the tools have not yet been integrated into our ACVC Classifier.RDRTec will exploit our current database containing ship structural features by incorporating aspect-dependency into our classification technique. We will explore the use of additional physics-based features for near-broadside aspect angles. In this work we will focus on features such as bow/stern shape and the presence of helicopter decks which have been identified by trained operators as useful for classification.