Separating and Sparse Imaging of GPR Diffractions by Dictionary Learning and Least-Squares Migration

Summary To efficiently separate weak diffractions from the GPR data, which usually has a single coverage and is easily contaminated with noise, we formulate a GPR diffraction separation method by incorporating the plane-wave destruction method and online dictionary learning technique. To promote the focusing ability of diffractions, a reweighted L2-norm and L1-norm minimization model is also introduced for accomplishing high-resolution GPR images, which has potential in focusing diffractions and reducing migration noise. The results obtained in our provided field example illustrates its good performance in separating and imaging of GPR diffractions and its potential value in illuminating fractures and the broken conditions.