Eigen-Template-Based HRR-ATR with Multi-Look and Time-Recursion

Automatic target recognition (ATR) using high range resolution (HRR) radar signatures is developed using classical Bayesian multiple hypothesis theory. An eigen-template-based matched filtering (ETMF) algorithm is presented where the templates are formed using the dominant range-space eigenvector of detected HRR training profiles and classification is performed using normalized matched filtering (MF). The proposed approach is extended to multi-look and sequential ATR where new observation profiles are recursively combined probabilistically with previous steps to update ATR results, which is useful for simultaneous recognition and tracking of moving targets. An HRR-specific profile normalization scheme is presented to satisfy matched filter requirements. Classification performance of the proposed method has been compared with a linear least-squares method and hidden Markov model (HMM) approach using MSTAR data collection.