A subspace decomposition technique to improve GPR imaging of anti-personnel mines

Ground-reßected clutter is often a performance-limiting factor in ground-penetrating radar (GPR) detection of nearsurface targets including anti-personnel mines. When a down-looking antenna is scanned across the surface this reßection produces a strong band in the image, which obscures shallow targets. Imperfections in the system impulse response (e.g., antenna ringing and cable reßections) can produce similar bands. Radar images of buried targets can be degraded by these forms of clutter. A multi-step process is described to improve imaging of near-surface objects. After range compression of the transmitted pulse, an estimate of the ground-reßected signal is obtained. Slow variations with scan position in the (complex) amplitude and time delay of this reßection are estimated in overlapping piecewise segments using a polynomial model. The resulting reßection estimate is subtracted from the data. Residual clutter is further reduced by using a subspace decomposition technique based on the generalized singular value decomposition (GSVD). In effect, the data are projected onto a space that is orthogonal to the reßection. Both surface reßections and ringing in the system impulse response are removed in this manner. The resulting cleaned data are then used with a conventional imaging algorithm (migration). The procedure is illustrated with data acquired by an OSU-developed GPR at the US Army Fort A.P. Hill, Site 71A.