Subsurface Target Localization by Single-View/Multistatic MUSIC Algorithm

The problem of detecting and localizing point-like targets buried in a half-space medium from near scattered field measurements under a single-view/multistatic configuration is addressed. In order to beat the resolution limits achievable by classical imaging methods, a MUSIC algorithm is employed. The MUSIC algorithm is known for providing super-resolving localization when the data correlation matrix is full rank. The point is that the latter circumstance does not hold true for the case at hand. This drawback can be remedied by adopting some smoothing procedure. However, for near-field configurations common smoothing strategies cannot be exploited since the data matrix lacks the required Vandermonde structure. To cope with this issue, the MUSIC algorithm is properly complemented by a suitable pre-processing stage which allows restoring the rank of the data matrix. Then, standard MUSIC algorithm is run and the achievable performance assessed via numerical examples for a 2D scalar geometry.