MAP image reconstruction for landmine and IED detection using ground penetrating radar

Abstract

ABSTRACTIn this article, we develop an iterative maximum a posteriori (MAP) estimation algorithm for reconstructing sub-surface images from ground penetrating radar (GPR) data. Note, the larger goal of our research is to use the resulting GPR images, along with appropriate detection algorithms, to identify landmines and improvised explosive devices. A typical scene-of-interest is expected to contain only a few electromagnetic wave scatterers. Taking advantage of this fact, we construct a novel sparsity promoting probability density function, which we refer to as the Butterworth prior because it is based on the well-known Butterworth lowpass filter and use it as the prior probability density function for the unknown reflection coefficients. The Butterworth prior can be used to approximate the uniform distribution but has the added benefit of mathematical tractability because its derivatives exist everywhere. Unlike available MAP-based approaches for reconstructing GPR images, we have developed a method for estimating the parameters of the Butterworth prior directly from the data. Therefore, at a cost of greater computational complexity, the main advantage of the proposed algorithm is that it does not require any user-defined parameters. We have tested the proposed algorithm, which we refer to as the MAP-Butterworth algorithm, on simulated and real data provided by the US Army Research Laboratory in Adelphi, MD. The results from the tests indicate that the MAP-Butterworth algorithm significantly suppresses sidelobes and background noise while retaining known scatterers.