A classification technique combining aspect dependence and elastic properties of target scattering

Abstract

Discrimination between man-made and natural underwater objects and between man-made objects of different characteristics are the key objectives of target classication. The current approach is mainly based on the analysis of the target signature imaged by high resolution (typically < 20 cm) high frequency (typically > 100 kHz) sonars. To estimate the potential of alternative classification schemes based on more detailed acquisition of target echo features, SACLANTCEN has investigated low frequency (10 kHz) sonar systems with high fractional (or relative) bandwidth (2-20 kHz). The classification method relies on two features of the target echo: aspect dependence and elastic scattering. The two dimensional reflection map, reconstructed from multiple aspects, serves as the basis for pre-classification. For objects of external shape recognized as man-made at the preclassification stage, resonance information is extracted by autoregressive spectral estimation techniques and further processed for particular aspects (e.g. broadside). The analysis of the resonance scattering provides an estimate of some geometrical and physical target parameters (i.e. shell wall thickness and material, and inner fluid properties) and can be improved by introducing the object dimensions, estimated from he reflection map. This report, which describes the classification methodology and the results obtained with steel cylinders, demonstrates the potential of the method.