Landmine detection using a ground penetrating radar

Reliable detection and clearing of landmines is still an unsolved problem. For solving the detection problem we performed measurements of inert mines in real environment using a commercial ultrawideband ground penetrating radar (GPR). Based on this data we developed signal processing strategies concerning the preprocessing of the data, and the target detection. Because of the severe clutter problem a special detection algorithm based on a Generalized Likelihood Ratio Test (GLRT) [1] was implemented featuring a signal to clutter improvement. For this article the statistical distribution of the GLRT-variable representing only clutter and noise was derived based on measured data. Now the complete system can be used as a CFAR detector. Furthermore by evaluating the detection rate for different environments it was possible to establish the ROC curves for the different scenarios. This can be used as a figure for comparing different mine detection systems. Applying a classifier the false alarm rate can be further reduced. By simply taking adjacent spatial alarms a robust classifier for bigger objects like anti-tank-mines could be developed. The resulting false alarm rate for near 100% detection is now acceptable. Classification of smaller anti-personal mines turns out to be a more complicated problem. Classification algorithms are under development.