Robust Multiping Forward-Looking Sonar Coherent Detection

Abstract

We tackle the problem of adaptive multiping detection of quasi-static point-like targets by moving forward-looking sonar based on the principles of coherent imaging. Data are first repositioned using an amplitude and phase coregistration procedure, enabling correct compensation of sonar displacements induced by the platform's movements. The result is a data stack similar to the space time adaptive processing case, in which the target observations are brought back into coherence, and the covariance matrix of the clutter naturally has an underlying structure. We then show that a rank-1 multiping detector incorporating a structured estimator of the covariance matrix performs significantly better than a conventional single-ping detector. However, while rank-1 multiping detection works well for ideally phase-aligned signals, this is not always the case in practical applications. Subspace detection, particularized to phase uncertainties, then leads to a new detection test robust to the target phase shift and whose performance still outperforms conventional approaches. Numerous simulations and analyses of experimental sonar data are discussed and illustrated.