Postprocessing of the Linear Sampling Method in Inverse Electromagnetic Scattering Problem for Obstacles

The linear sampling method is known to be a simple and computationally efficient approach to retrieve the support of the scatterer using multistatic scattered field data. However, the recovered profile is always misleading, owing to the lack of robust edge detecting. This paper addresses this open issue. Using moving least square approximation, the upper and lower bounds of the profile of scatterers are pursued, and a sweeping process finds the optimal profile to match the scattered field data.