Compressive Sensing based Software Defined GPR for Subsurface Imaging

Abstract

This paper presents the design of a new ground penetrating radar (GPR) integrating software defined radio (SDR) and compressive sensing (CS) technologies. In recent literature, SDR has been explored for designing GPR in the way of stepped-frequency continuous wave (SFCW) radar. In the operation, the software defined GPR (SD GPR) radiates a series of sinusoidal signals of evenly spaced frequencies at each scan position. The reflection signals of all frequency tones are received and their amplitude and phase responses are measured and characterized. As each individual frequency tone needs to be generated, transmitted and received in sequence, it results in a slow scan speed. In this study, compressive sensing is explored to expedite SD GPR operation speed. For SD GPR subsurface survey, when the target area is spatially sparse, i.e. the buried objects are sparsely distributed, In addition, we develop a CS based signal processing algorithm specifically for improving image quality and reducing the clutter. To reconstruct the image correctly, an automatic parameter selection algorithm based on the structural similarity index measure (SSIM) is proposed. For validation, a laboratory test was conducted. The experimental results demonstrate that the CS imaging algorithm produces less clutter comparing with the traditional GPR image algorithms, such as the time domain back projection (BPA) method.