Abstract
In this paper, a 1-D Eigen-Template based Matched Filtering (ETMF) algorithm is proposed for High Range Resolution (HRR) radar signatures to classify ground targets. It is demonstrated that effective HRR ATR performance can be achieved if the templates are formed via Singular Value Decomposition (SVD) of detected HRR profiles and the classification is performed using normalized Matched Filtering (MF). It is also shown theoretically that the eigen-vectors are the optimal feature set representation of a collection of HRR profile vectors. The SVD operation is also shown to produce orthogonal range and angle basis space vectors in a convenient decoupled form and it is proposed that the dominant range-space eigenvector be used as templates. The proposed approach is then extended to perform multi-look and sequential ATR where, recognition at previous steps are used along with new observation profiles to update ATR results which is appropriate for simultaneous recognition and tracking of moving targets. Simulation results comparing the proposed approach with an existing baseline linear least-squares method are presented for nine-target MSTAR data set. The effect of applying Power Transform (PT) operation on noisy observation profiles is also studied.