Applying Blind Source Separation to Magnetic Anomaly Detection Algorithms

Magnetic anomaly detection (MAD) is highly effective for detecting unexploded ordnance in marine environments. There is a need to localize multiple targets simultaneously when their magnetic signatures mix. Independent component analysis (ICA) has been used as a method for blind source separation in biomedical fields to separate various electrical and magnetic field signals associated with the human body. By using multiple magnetometers on single or multiple vehicles, ICA can be adapted to separate out the overlapping signals for a magnetic survey. However, due to scaling and permutation ambiguities associated with the ICA process, the separation step results in errors during the localization step. These issues are mitigated so that successful signal separation and localization becomes possible. According to simulations, the ICA algorithm combined with a genetic algorithm (GA) is able to produce localizations for two magnetic targets on the ocean floor within 1 to 2 meters of accuracy.