A novel signal processing technique for clutter reduction in GPR measurements of small, shallow land mines

The global land mine crisis is creating immense social and economic problems worldwide. Safe and cost effective methods for clearing these mines are therefore needed. A promising sensor for detecting small anti-personnel mines is the ground penetrating radar (GPR). However, GPR performs inadequately due to clutter which dominates the return and obscures the mine information.; In this dissertation a signal processing technique is developed which can be used to reduce clutter in GPR data. It is assumed that the GPR system is designed to reduce the system clutter by calibration and also that it is operated to reduce the time duration of the clutter by using normal incidence angles. The new preprocessing employs a parametric modeling approach for clutter reduction. Frequency domain basis functions are developed to represent the clutter and the mine contributions in the GPR data. An iterative signal processing technique is developed to estimate the unknown parameters in the basis functions and reduce the clutter. The new technique also improves existing signal processing techniques by incorporating an adaptive basis function for clutter representation. The new algorithm is robust to the variability of clutter between measurements and accounts for the uncertainty in the GPR clutter characteristics. The facts that land mines are buried at shallow depths, that their returns are small relative to that of the clutter are compensated for in the new processing technique. The returns from subsurface inhomogeneities, for example rocks and tree roots are also treated as part of the clutter.; To assess the performance of the clutter reduction technique, the cross correlation between a reference signature of the target of interest and the clutter reduced data is used. To quantify the performance of clutter reduction, improvement in detection is assessed though the detector receiver-operating-characteristic (ROC) curves. A simple matched filter, formulated to account for the uncertainty in the placement of the mine is used as a detector and it is preceded by the clutter reduction preprocessing. Several simulated and measured data sets are considered. A two-dimensional finite-difference-time domain (FDTD) code is implemented to provide a controlled environment for generating simulated data. The experimental GPR data is acquired at Fort A. P. Hill, VA using a dielectric rod antenna.; A multi spectral approach is also developed to discriminate between similar targets. It exploits the frequency sub band information available in the mine signatures and is applied at a single look angle of the antenna.