| Abstract [eng] |
This research analyzes algorithms for automating the detection of explosives. Three methods were analyzed: local outlier factor (LOF), wavelet transformation, and amplitude of the analytic signal (AAS) algorithms. The study discusses the strengths and weaknesses of all three methods, applying these methods to synthetic data, real magnetometer data from scans in Lithuania, and magnetometer data from an online source. The results of the study showed that the effectiveness of the algorithms depends on the magnetometer used and the selected algorithm parameters. In the case of real scans in Lithuania, the best results were shown by the amplitude of the analytic signal algorithm, which identified 71$\%$ of the explosive locations. However, none of the algorithms were able to identify all neutralized explosives during scans in Lithuania, so it is recommended that the algorithms be used only as auxiliary tools for searching for explosive devices. |
