Sparse Ultra Wide Band Radar imaging in a locally adapting matching pursuit (LAMP) framework

Ultra-Wide Band (UWB) Impulse Radar imaging finds widespread applications in extending human vision beyond media that cannot be penetrated by visible light. However, high resolution imaging demands acquisition of large amounts of data, which has motivated researchers towards the application of Compressed Sensing techniques in radar imaging, if the target space is sparse. It might be noted that apart from sparsity, most UWB radar signatures also exhibit another property of being clustered into groups that conventional sparse reconstruction techniques like Orthogonal Matching Pursuit (OMP) or Lasso rarely take into account. Though attempts have been made to reconstruct block sparse signals using BlockOMP or Group Lasso, these algorithms inherently assume specific locations and sizes of the possible clusters, and hence are not satisfactory if the clusters are randomly located and variably sized. In this manuscript, the traditional block based approach has been extended for Multiple Measurement Vector (MMV) model. To address its shortcomings, a novel framework that is referred to as a locally adapting matching pursuit (LAMP) has been proposed for the efficient reconstruction of radar signatures from compressed measurements considering both sparsity and clustering priors, but without assuming any specific group structure like BOMP or Group Lasso. The performance of the proposed LAMP has been compared with existing algorithms in different scenarios. The framework has also been successfully applied on a real-world radar imaging data, with different types of sensing matrices.