Landmine feature extraction in UWB SAR based on sparse representation

Ultrawide Band Synthetic Aperture Radar (UWB SAR) is an alternative to detect landmines. The echo of a landmine has “double-hump” signature, which corresponds to the returns of front and rear edges of the top of the landmine. However, their echo traces do not fit with the corresponding integral traces of the SAR imaging model, which can lead to defocusing in the SAR image. In this paper, we construct two dictionaries for the front peak and rear peak, respectively. Then we find the optimally sparse representation of each peak of the double-hump signature via basis pursuit algorithm. The results of the real data experiment show the validity of the method.