Optimized SVD Approach for the Detection of Weak Subsurface Targets From Ground-Penetrating Radar Data

We propose a two-pass ground-penetrating radar (GPR) approach to detect recently buried nonmetallic scatterers whose signals are otherwise undistinguishable from surrounding clutter. The operating framework is based on a pass-1 data collection over an object-free ground and a pass-2 data collection over the same ground, yet at unregistered positions, which may be contaminated by a recently implanted dielectric scatterer. The detection algorithm is based on a modified singular value decomposition approach and is shown to perform significantly better than a direct signal cancellation approach. Validating results are presented on measured target-free data to which synthetic targets of various strengths are added. Multiple realizations along the measurement path of the GPR array provide statistical diversity to the results which is cast as receiving operating curves, showing quantitatively the improved detection threshold of the new method.