The application of fusion of multiple aspect scattered data to PMUSIC-vector based electromagnetic target identification

This paper presents an electromagnetic target identification method in the resonance scattering region, which uses pseudospectrum multiple signal classification (PMUSIC) vectors as feature vectors. These feature vectors are obtained with the fusion of multiple aspect scattered fields of the targets into one covariance matrix for each target. The proposed method initially stores the suitable late-time interval of the scattered fields of a target obtained at different aspect angles. Then, the correlation information of these signals is gathered into one covariance matrix by a simple fusion technique. By using this covariance matrix and PMUSIC algorithm, feature vectors of the targets are extracted and stored to database. In the test phase, PMUSIC algorithm is applied to the same suitable late-time interval of test field (signal) belonging to a test target and test (PMUSIC) vector is evaluated. Finally, the decision about the test target is given according to the highest correlation coefficient between feature vectors and test vector. The proposed method is demonstrated for small-scale airplane targets modeled by thin wires. It is shown that in addition to its short runtime, the method gives satisfactory accuracy rates even for low signal-to-noise (SNR) ratios.