Ultrasound NDE imaging through reverberant layers via subspace analysis and projection

In ultrasonic flaw imaging through reverberant layers, the top layer not only induces strong reverberation clutter but also repeats the flaw echo, giving rise to echo multiples. The reverberation clutter often overwhelms the target echoes, rendering the imaging of material flaws extremely difficult. Furthermore, echo multiples may obscure the identification of actual targets, especially when there are a number of targets in the scene. The reverberation clutter and flaw echo multiples must be suppressed or sufficiently mitigated in order to enable proper flaw imaging. In this paper, a subspace-based approach is utilized to suppress the reverberation clutter, followed by predictive filtering to suppress echo multiples. The clutter, due to its significantly high strength relative to the flaw echoes, is captured in the signal subspace that corresponds to the dominant eigenvalues of the data covariance matrix. It can be suppressed by applying orthogonal clutter subspace projection. Predictive deconvolution method is then used to suppress flaw echo multiples and enhance the visibility of the true flaw echoes in the presence of noise and clutter remnants. The effectiveness of the proposed imaging-through-layer approach is demonstrated via synthesized array data using real measurements of reverberation clutter and flaw echoes.