Multi-feature based landmine identification using ground penetrating radar

In this paper a method is presented to identify landmines in various burial conditions through features obtained from two analytic methods, PCA and DFT. A GPR data set is processed to obtain landmine data. For this process, signal regulation, average filtering and the ground effect reduction are performed. After the landmine data is obtained, PCA and DFT are applied for extracting features of each landmine. A database is then constructed based on landmine features for different levels of moisture and types of ground, and burial depths of the landmines. Using the database, identification is performed by comparing features against the database. For the identification, a counting method is used. If the features of an input GPR data match those of a landmine stored in the database, the counting value for that landmine is increased. The landmine with the largest counting value is given as the identified landmine. Four landmines are considered buried at different depths in various ground conditions. Examples show that the proposed method is capable of landmines efficiently.