Two-step (estimate and detect) sparse imaging

We propose a two-step algorithm, for electromagnetic imaging of perfect electric conducting (PEC) targets hidden inside known medium, that uses l₁ regularization. The basic assumption is that the target can be approximated by a few equivalent sources (filament currents). In the first step, for each incidence, we estimate the locations of the equivalent sources corresponding to the target scattering centers. However, false solutions may appear due to the multiple scattering or insufficient measurements. We eliminate the false solutions in the second step; we decide for each of the estimated equivalent sources weather it belongs to the target.