Landmine feature extraction in GPR(SFCW) data

This paper shows an algorithm based on linear geostatistics in ground penetrating radar (GPR) data for extracting landmine features. That is a major problem for civilian population in different rural areas in Colombia. Because linear geostatistics has a main difference respect to classic statistics, this technique can be used to detect, to create a signal with object information, and to extract information about physical parameters of buried object such as a landmine as Improvised Explosive Device (IED). Consecutive acquired signals provide a correlated relationship useful for both clutter reduction (background and test lane limits) by using Principal Component Analysis (PCA) and separation of object information by using geostatistics. The samples of resulting signal are compared with a threshold value to obtain a signal with features related to position and object size. The experimental data have been collected by using frequency stepped GPR system covering a range from 700[MHz] to 10[GHz] in a Military Base with real conditions.